Modelling Financial Markets

Chapter 20 (special chapter): Path Signatures and Realised Volatility — a Research Case Study

Chapter 20 (special chapter): Path Signatures and Realised Volatility — a Research Case Study

Daniel Traian PELE

Bucharest University of Economic Studies
IDA Institute Digital Assets
Blockchain Research Center
AI4EFin Artificial Intelligence for Energy Finance
Romanian Academy, Institute for Economic Forecasting
MSCA Digital Finance

Signature-based techniques give mathematical insight into the interactions between complex streams of evolving data.

— Lyons & McLeod (2022), Signature methods in machine learning

Gu et al. (2024) describe a recent window of a series as a path and summarise it by its path signature. Signature features enter a two-step LASSO (Least Absolute Shrinkage and Selection Operator) regression, and past periods whose path resembles the current one receive a larger weight through a signature kernel. For freight rates, the authors report a more than fivefold accuracy gain over an industry forecast. The chapter asks whether the same method improves forecasts of daily realised variance (RV), where the benchmark is the heterogeneous autoregressive (HAR) model of Corsi (2009) and, above all, its log version, log-HAR.

The transfer is pre-registered: four hypotheses are fixed before the results, and the data are 50 assets from the VOLARE (VOLatility Archive for Realized Estimates, database) archive, 40 of them stocks. Each component of the method is tested separately, so that any gain can be attributed either to the features or to the weights. The comparisons use the QLIKE (quasi-likelihood) loss, asset-by-asset Diebold–Mariano tests and the Holm and Benjamini–Hochberg corrections for multiple testing.

The result is null. At the one-day horizon, Sig-L, the model with signature features, has a mean QLIKE loss 0.4% higher than log-HAR. Sig-L is significantly better than log-HAR on 3 of 50 assets without correction and on a single one after the Holm correction, while about 2.5 raw rejections are expected by chance alone. At one week and one month, the full model, Sig-LK, has a QLIKE loss 28.0% and 92.9% higher than log-HAR. The strongest effect in the study is the oldest one: log-HAR is significantly better than HAR in levels on 21 of 50 assets at one day, without correction. The chapter shows, through ablations and the mechanism, why the method does not transfer.

Ratio of each model's mean QLIKE loss to that of the HAR model, for each of the 50 assets; a ratio below 1 means a more accurate model than HAR. The boxes span the interquartile range across assets, the line in the box is the median and the diamond is the mean.
Ratio of each model's mean QLIKE loss to that of the HAR model, for each of the 50 assets; a ratio below 1 means a more accurate model than HAR. The boxes span the interquartile range across assets, the line in the box is the median and the diamond is the mean. Interactive chart: “QLIKE Relative to HAR, Asset by Asset” →

What you will be able to do

  • Compute the signature of a piecewise-linear path and its Lévy area.
  • Compute signature features and kernel weights for a volatility forecast.
  • Interpret forecast comparisons across many assets, with Diebold–Mariano tests, model confidence sets and Holm and Benjamini–Hochberg corrections.
  • Interpret the result of a pre-registered transfer to a new domain, including a null result together with its mechanism.

The Further Topics and the Appendix are reading for familiarisation.

Chapter plan

  1. Route and MotivationThe chapter's question, the two components of the method and the HAR benchmark.
  2. Realised Measures as Forecast TargetsRealised variance, its measurement error and the loss functions.
  3. The VOLARE DatabaseThe panel of 50 assets in three classes.
  4. Benchmarks: the HAR FamilyHAR, log-HAR and the seven models compared.
  5. Path SignaturesThe definition, Chen's identity, the Lévy area and the signature kernel.
  6. The Paper: Gu et al. (2024)The freight-rate problem, the method and the limits of the evidence.
  7. The Transfer Design (Pre-Registered)The four hypotheses, rolling estimation and the evaluation protocol.
  8. ResultsQLIKE comparisons, asset-by-asset DM (Diebold–Mariano, test) tests, model confidence sets and panel tests.
  9. Ablations and RobustnessThe 2 × 2 ablation and results by asset class.
  10. Limitations and Open DirectionsWhat the study cannot claim and what remains to be tested.
  11. Case Study: The Signature as a Lead-Lag MetricSPY (SPDR S&P 500 ETF Trust) and Bitcoin at 5 minutes: who moves first.
  12. AI for Scientific DiscoveryWhen does a method transfer from one domain to another.
  13. ConclusionsKey ideas and the practical decision.
  14. Further TopicsDetailed examples and advanced topics.
  15. AppendixFeatures, estimation, evaluation periods and checks.

Route and Motivation

The chapter tests whether a forecasting method published for freight rates improves forecasts of daily realised variance. The method has two ingredients, signature features of a recent data window and kernel weights that favour similar past windows, and each is tested on its own against the HAR family of benchmarks. The transfer is pre-registered: the hypotheses, the panel and the tests are fixed before the results are seen.

Does a Method Built for Freight Rates Improve Volatility Forecasts?

Slide 2 in the PDF

The method of Gu et al. (2024) was developed to forecast freight rates at Amazon. The chapter asks whether it also improves forecasts of realised variance, using 50 assets from VOLARE (VOLatility Archive for Realized Estimates) (Cipollini et al., 2026). The question is one of transfer: a method that works in one domain is moved to another, with a design fixed in advance.

The chapter develops four skills:

  1. To compute the signature of a piecewise-linear path and its Lévy area.
  2. To compute signature features and kernel weights for a volatility forecast.
  3. To interpret forecast comparisons on many assets, with Diebold–Mariano (DM) tests, model confidence sets and the Holm and Benjamini–Hochberg (BH) corrections.
  4. To interpret the result of a pre-registered transfer, including a null result and its mechanism.

Detailed examples are in the Further Topics, after the Conclusions.

In brief

Path Shape and Volatility Forecasts
  • A research case study: the method of Gu et al. (2024), transferred to daily realised variance
    • the method was published for freight rates at Amazon
    • data: 50 assets from VOLARE (VOLatility Archive for Realized Estimates) Cipollini et al. (2026)
  • Detailed examples and advanced topics are in Further Topics, after the Conclusions
    • the footer button leads there; the Back button returns to the starting slide
What you should be able to do
  • Compute the signature of a piecewise-linear path and its Lévy area
  • Compute signature features and kernel weights for a volatility forecast
  • Interpret forecast comparisons on many assets
    • with Diebold–Mariano (DM) tests and model confidence sets
    • with Holm and Benjamini–Hochberg (BH) corrections
  • Interpret the result of a pre-registered method transfer, including a null result and its mechanism
  • Further Topics and Appendix: reading for familiarity

Why Forecast Realised Variance?

Slide 3 in the PDF

Volatility is an input to almost every risk measure of the course: tail-risk measures in Chapter 7, option prices in Chapter 12 and portfolio weights in Chapter 4. Intraday data make the daily variance almost observable through realised variance (RV), the measure of Chapter 9. Volatility forecasting therefore becomes a regression problem with a noisy but model-free target.

The benchmark is hard to beat. The HAR (heterogeneous autoregressive) model of Corsi (2009) uses three averages of past RV, is linear and is estimated by ordinary least squares (OLS) in a second. A new method has to beat HAR out of sample, on many assets and with a proper test, before it can be said to improve volatility forecasts.

In brief

  • Volatility is an input to almost every risk measure in this course
  • Intraday data make daily variance almost observable
    • the measure is realised variance (RV, Chapter 9)
    • So volatility forecasting becomes a regression problem with a noisy but model-free target
  • The benchmark is hard to beat
    • the HAR (heterogeneous autoregressive) model Corsi (2009) (Chapter 9)
    • Three averages of past RV; linear; estimated by OLS (ordinary least squares) in a second
  • Any new method must beat HAR out of sample, on many assets, with a proper test

Two Ingredients to Test: Signature Features and Kernel Weights

Slide 4 in the PDF

Gu et al. (2024) describe a recent window of a series as a path and summarise it by its signature. The signature features enter a sparse linear regression, a LASSO (least absolute shrinkage and selection operator) estimated in two steps. Past periods whose path looks like the current one receive more weight through a signature kernel. For freight rates the authors report a more than fivefold accuracy gain over an industry forecast.

The chapter's question is whether the same recipe improves on HAR-type models for realised variance. Because the method combines two ingredients, each is tested separately: a gain can then be attributed either to the features or to the weights, and a loss located in one of them.

In brief

  • Gu et al. describe a recent window of a series as a path and summarise it by its signature
    • Signature features enter a sparse linear regression (LASSO, least absolute shrinkage and selection operator, in two steps)
    • Past periods whose path looks like today’s get more weight (a signature kernel)
  • They report a more than fivefold accuracy gain over an industry forecast of freight rates
    • Question: does the same recipe improve on HAR-type models for realised variance?
  • Each ingredient is tested separately, so that a gain can be attributed to the features or to the weights
Two Ingredients to Test: Signature Features and Kernel Weights Wall Street and the New York Stock Exchange, 25 March 2020 Photo: Billie Grace Ward (2020); CC BY 2.0; Wikimedia Commons

Realised Measures as Forecast Targets

Before any model is compared, the target must be defined. This section states the price model behind realised variance, the measurement error of RV and the auxiliary measures in the VOLARE data, and then fixes the forecast target, the alignment of forecasts with targets and the loss functions used to rank the models.

Realised Variance and Quadratic Variation (1/2)

Slide 5 in the PDF

The question is what exactly is forecast, and how intraday data measure the variance of a day. The log price \(p_t\), with time in days, moves by a drift, a diffusion term and jumps:

\[dp_t = \mu_t\,dt + \sigma_t\,dW_t + dJ_t,\]
(20.1)

where \(\mu_t\) is the drift, \(\sigma_t\) the spot volatility, \(W_t\) a Brownian motion and \(J_t\) a pure-jump process, with \(\Delta J_s\) the jump size at time \(s\).

Realised variance is the sum of squared intraday returns of day \(t\) (Barndorff-Nielsen & Shephard, 2002):

\[\mathrm{RV}_t = \sum_{i=1}^{M} r_{t,i}^2 \;\xrightarrow{\;M\to\infty\;}\; \int_{t-1}^{t}\sigma_s^2\,ds + \sum_{t-1<s\le t} (\Delta J_s)^2,\]
(20.2)

where \(M\) is the number of intraday returns in day \(t\) and \(r_{t,i}\) the \(i\)-th intraday log return; for a US (United States) stock, \(M = 78\) five-minute returns cover the 6.5-hour session. The limit in (20.2) is the quadratic variation of the day: the integrated variance plus the sum of squared jumps.

In brief

  • Question: what exactly do we forecast, and how do intraday data measure the variance of a day?
  • The price model: the log price moves by a drift, a diffusion term and jumps
    \begin{equation*} dp_t = \mu_t\,dt + \sigma_t\,dW_t + dJ_t \end{equation*}
    • \(p_t\): log price at time \(t\) (time in days); \(\mu_t\): drift; \(\sigma_t\): spot volatility; \(W_t\): Brownian motion; \(J_t\): pure-jump process, \(\Delta J_s\) = jump size at time \(s\)
  • Realised variance: the sum of squared intraday returns of day \(t\) Barndorff-Nielsen & Shephard (2002)
    \begin{equation*} \RV_t = \sum_{i=1}^{M} r_{t,i}^2 \;\xrightarrow{\;M\to\infty\;}\; \int_{t-1}^{t}\sigma_s^2\,ds + \sum_{t-1<s\le t} (\Delta J_s)^2 \end{equation*}
    • \(M\): number of intraday returns in day \(t\) (for a US stock, 78 five-minute returns in the 6.5-hour session); \(r_{t,i}\): the \(i\)-th intraday log return of day \(t\)
    • The limit is the quadratic variation of day \(t\): the integrated variance plus the sum of squared jumps

Realised Variance and Quadratic Variation (2/2)

Slide 6 in the PDF

Without jumps and noise, on an equally spaced grid, the limit of RV is the integrated variance \(\mathrm{IV}_t = \int_{t-1}^{t}\sigma_s^2\,ds\), the true variance of day \(t\), and the error is mixed Normal:

\[\sqrt{M}\,(\mathrm{RV}_t - \mathrm{IV}_t) \Rightarrow \mathrm{MN}\big(0,\; 2\,\mathrm{IQ}_t\big), \qquad \mathrm{IQ}_t = \int_{t-1}^{t}\sigma_s^4\,ds,\]
(20.3)

where \(\mathrm{IQ}_t\) is the integrated quarticity and \(\mathrm{MN}(0, V)\) a Normal law with a random variance \(V\): given the volatility path, the error is Normal with variance \(2\,\mathrm{IQ}_t/M\).

RV is therefore a noisy measurement of \(\mathrm{IV}_t\), and the noise is largest on volatile days, where \(\mathrm{IQ}_t\) is large. This measurement error is what HARQ, the HAR model with realised quarticity, corrects for. In the other direction, microstructure noise makes very fine sampling biased, and five-minute returns are the standard compromise (Chapter 9).

In brief

  • Without jumps and noise, on an equally spaced grid, the limit is the integrated variance
    • the error is mixed Normal
    \begin{equation*} \sqrt{M}\,(\RV_t - \mathrm{IV}_t) \Rightarrow \mathrm{MN}\!\left(0,\; 2\,\mathrm{IQ}_t\right), \qquad \mathrm{IV}_t = \int_{t-1}^{t}\sigma_s^2\,ds, \quad \mathrm{IQ}_t = \int_{t-1}^{t}\sigma_s^4\,ds \end{equation*}
    • \(\mathrm{IV}_t\): integrated variance, the true variance of day \(t\); \(\mathrm{IQ}_t\): integrated quarticity
    • \(\mathrm{MN}(0, V)\): mixed Normal, a Normal with a random variance \(V\); given the volatility path, the error is Normal with variance \(2\,\mathrm{IQ}_t/M\)
  • Reading: RV is a noisy measurement of \(\mathrm{IV}_t\)
    • the noise is largest on volatile days (where \(\mathrm{IQ}_t\) is large)
    • This is the measurement error that HARQ, the HAR model with realised quarticity, corrects for (HAR section)
  • Microstructure noise makes very fine sampling biased
    • five-minute returns are the standard compromise (Chapter 9)

Semivariances and Quarticity in VOLARE

Slide 7 in the PDF

Realised semivariances (Barndorff-Nielsen, Kinnebrock & Shephard, 2010) split RV by the sign of the intraday returns: \(\mathrm{RV}_t^{+} = \sum_i r_{t,i}^2\mathbf{1}\{r_{t,i}>0\}\) is the variance from up-moves and \(\mathrm{RV}_t^{-} = \sum_i r_{t,i}^2\mathbf{1}\{r_{t,i}<0\}\) the variance from down-moves, and they add up to \(\mathrm{RV}_t\).

Realised quarticity estimates \(\mathrm{IQ}_t\), the scale of the measurement error of RV:

\[\mathrm{RQ}_t = \frac{M}{3}\sum_{i=1}^{M} r_{t,i}^4 \;\to\; \mathrm{IQ}_t.\]
(20.4)

The factor \(M/3\) comes from the Normal law: a return with variance \(s^2\) has fourth moment \(3s^4\), and each five-minute variance is about \(\mathrm{IV}_t/M\). The limit requires a continuous, noise-free price and an equally spaced grid; a jump \(\Delta J\) adds about \(M(\Delta J)^4/3\), so raw RQ (Realised Quarticity) is sensitive to jumps.

In brief

  • Semivariances Barndorff-Nielsen, Kinnebrock & Shephard (2010): RV split by the sign of the intraday returns
    \begin{equation*} \RV_t^{+} = \sum_{i=1}^{M} r_{t,i}^2\,\mathbf{1}\{r_{t,i}>0\}, \qquad \RV_t^{-} = \sum_{i=1}^{M} r_{t,i}^2\,\mathbf{1}\{r_{t,i}<0\} \end{equation*}
    • \(\RV_t^{+}\): variance from up-moves, \(\RV_t^{-}\): variance from down-moves
    • They add up: \(\RV_t = \RV_t^{+} + \RV_t^{-}\)
  • Realised quarticity (RQ): an estimate of \(\mathrm{IQ}_t\), the scale of the measurement error of RV
    \begin{equation*} \mathrm{RQ}_t = \frac{M}{3}\sum_{i=1}^{M} r_{t,i}^4 \;\to\; \mathrm{IQ}_t \end{equation*}
    • The factor \(M/3\): a Normal return with variance \(s^2\) has fourth moment \(3s^4\), and each five-minute variance is about \(\mathrm{IV}_t/M\)
    • The limit needs a continuous, noise-free price and an equally spaced grid; a jump \(\Delta J\) adds about \(M(\Delta J)^4/3\), so raw RQ is jump-sensitive

Targets and Losses (1/2): the Target

Slide 8 in the PDF

The target at horizon \(h\) is the average RV over the next \(h\) days,

\[\mathrm{RV}_{t+1:t+h} = \frac{1}{h}\sum_{j=1}^{h}\mathrm{RV}_{t+j}, \qquad h \in \{1, 5, 22\},\]
(20.5)

where \(t\) is the forecast origin, the last day observed, and \(h = 1, 5, 22\) are one day, one week and one month of trading days.

Log models forecast \(y = \ln \mathrm{RV}_{t+1:t+h}\) and return to levels with a variance correction, \(F_t = \exp(\hat y_t + \hat s^2/2)\), where \(F_t\) is the forecast made at \(t\), \(\hat y_t\) the fitted log forecast and \(\hat s^2\) the variance of the regression residuals. The correction follows from \(\mathbb{E}[e^Z] = e^{\mu + s^2/2}\) for \(Z \sim N(\mu, s^2)\); it is exact if the log error is conditionally Normal with constant variance and an approximation otherwise.

For \(h > 1\) consecutive targets overlap, which induces serial dependence through lag \(h-1\) by construction and possibly longer. Every test therefore uses a HAC (heteroskedasticity and autocorrelation consistent) variance.

In brief

  • The target at horizon \(h\) is the average RV over the next \(h\) days
    \begin{equation*} \RV_{t+1:t+h} = \frac{1}{h}\sum_{j=1}^{h}\RV_{t+j}, \qquad h \in \{1, 5, 22\} \end{equation*}
    • \(t\): the forecast origin (the last day observed); \(h = 1, 5, 22\): one day, one week, one month of trading days
  • Log models forecast \(y = \ln\RV_{t+1:t+h}\) and return to levels with a variance correction
    \begin{equation*} F_t = \exp(\hat y_t + \hat s^2/2) \end{equation*}
    • \(F_t\): forecast of \(\RV_{t+1:t+h}\) made at \(t\); \(\hat y_t\): fitted log forecast; \(\hat s^2\): variance of the regression residuals
    • \(+\hat s^2/2\): since \(\E[e^Z] = e^{\mu + s^2/2}\) for \(Z \sim N(\mu, s^2)\); exact if the log error is conditionally Normal with constant variance, otherwise an approximation
  • For \(h > 1\) targets overlap: this induces serial dependence
    • through lag \(h - 1\) by construction, possibly longer
    • so every test uses a HAC (heteroskedasticity and autocorrelation consistent) variance

Matching a Forecast to Its Target

Slide 9 in the PDF

The subscripts play two roles. In the definition of the target, \(t\) is the forecast origin; in the loss and test formulas, the subscript labels a matched evaluation row, which pairs the realised target with the forecast made before that target became observable.

At the one-day horizon, information through Monday's close forecasts Tuesday's RV, and the forecast is stored next to the Tuesday realisation. At the five-day horizon the same origin forecasts the average from Tuesday to the following Monday, and the next origin gives an overlapping target. The forecast comparison tests use the same realised target for both models in each row. The row label is not permission to use that target in estimation.

In brief

  • In the target definition, the subscript \(t\) is the forecast origin
  • In loss and test formulas, the subscript labels a matched evaluation row
    • the realised target
    • the forecast made before that target became observable
  • One-day horizon: information through Monday close forecasts Tuesday RV
    • stored next to the Tuesday realisation
  • Five-day horizon: the same origin forecasts the average from Tuesday to the following Monday
    • the next origin gives an overlapping target
  • Forecast comparison tests use the same realised target for both models in each row
  • Sharing a row with the forecast does not allow the target to be used in estimation

Targets and Losses (2/2): Robust Losses

Slide 10 in the PDF

A loss is robust in the sense of Patton (2011) if, for a conditionally unbiased proxy, the ranking of models by expected loss computed with \(\mathrm{RV}_t\) is the same as with the true variance. Conditional unbiasedness means \(\mathbb{E}_{t-1}[\mathrm{RV}_t] = \sigma_t^2\), where \(\mathbb{E}_{t-1}\) is the expectation given information up to \(t-1\) and \(\sigma_t^2\) the true variance of day \(t\).

The chapter uses two robust losses, written here for \(h = 1\) (for \(h > 1\), \(\mathrm{RV}_t\) is replaced by the target):

\[\mathrm{MSE}_t = (\mathrm{RV}_t - F_t)^2, \qquad \mathrm{QLIKE}_t = \frac{\mathrm{RV}_t}{F_t} - \ln\frac{\mathrm{RV}_t}{F_t} - 1 \;\ge\; 0.\]
(20.6)

MSE (mean squared error) is the squared error. QLIKE (quasi-likelihood loss) is zero only if \(F_t = \mathrm{RV}_t\) and grows faster when the forecast is too low than when it is too high: with \(\mathrm{RV}_t = 1\), \(F_t = 0.5\) gives \(2 - \ln 2 - 1 = 0.307\) and \(F_t = 2\) gives \(0.5 - \ln 0.5 - 1 = 0.193\).

QLIKE depends only on the ratio \(\mathrm{RV}_t/F_t\), so the same proportional error has the same loss on a calm day and on a stormy one, whereas MSE is dominated by a few crisis days. QLIKE is therefore the primary loss and MSE the secondary one.

In brief

  • A loss is robust Patton (2011)
    • for a conditionally unbiased proxy, the expected-loss ranking with \(\RV_t\) equals that with the true variance
    • Conditionally unbiased: \(\E_{t-1}[\RV_t] = \sigma^2_t\), with \(\E_{t-1}\) the expectation given information up to \(t - 1\) and \(\sigma^2_t\) the true variance of day \(t\)
  • The two robust losses used here (written for \(h = 1\); for \(h > 1\) replace \(\RV_t\) by the target)
    \begin{equation*} \mathrm{MSE}_t = (\RV_t - F_t)^2, \qquad \QL_t = \frac{\RV_t}{F_t} - \ln\frac{\RV_t}{F_t} - 1 \;\ge\; 0 \end{equation*}
    • MSE (mean squared error): the squared error; QLIKE (quasi-likelihood loss) is zero only if \(F_t = \RV_t\) and grows faster when the forecast is too low than when it is too high
    • Example with \(\RV_t = 1\): \(F_t = 0.5\) gives \(\QL_t = 2 - \ln 2 - 1 = 0.307\), \(F_t = 2\) gives \(0.5 - \ln 0.5 - 1 = 0.193\)
  • QLIKE depends on the ratio \(\RV_t/F_t\)
    • the same proportional error has the same loss on a calm day as on a stormy day
  • MSE is dominated by a few crisis days
    • QLIKE is the primary loss, MSE the secondary one

The VOLARE Database

The transfer is tested on a panel from VOLARE, which provides five-minute realised measures for many assets. The section describes the 50 assets in three classes, the variables used and the behaviour of realised volatility across classes.

The Panel: 50 Assets in Three Classes

Slide 11 in the PDF

The question is on which data the transfer is tested, and why on three asset classes. The panel has:

  • 40 American large-cap stocks, 2 January 2015 – 30 January 2026, with 111,440 observations and a median annualised volatility of 17.1%;
  • 5 futures, 28 September 2009 – 30 January 2026, with 21,059 observations and 19.7%;
  • 5 currency pairs, 25 September 2009 – 30 January 2026, with 21,203 observations and 7.3%.

The stocks are 40 members of the Dow Jones 30 and S&P 100 (Standard & Poor's), labelled by ticker, from AAPL (Apple) to XOM (Exxon Mobil). The futures are corn (C), crude oil (CL), the E-mini S&P 500 (ES), gold (GC) and natural gas (NG). The FX (foreign exchange) series are the Australian dollar, euro, British pound, Canadian dollar and Japanese yen against the US dollar.

In brief

ClassAssetsSampleObservationsMedian
volatility
American large-cap stocks402 Jan 2015 – 30 Jan 2026111 44017.1%
Futures (C, CL, ES, GC, NG)528 Sep 2009 – 30 Jan 202621 05919.7%
FX (AUD, EUR, GBP, CAD, JPY vs USD)525 Sep 2009 – 30 Jan 202621 2037.3%
  • Question: on which data do we test the transfer, and why on three asset classes?
  • Assets
    • Stocks: 40 members of the Dow Jones 30 and S&P 100 (Standard & Poor's), labelled by ticker, from AAPL (Apple) to XOM (Exxon Mobil)
    • Futures: C = corn, CL = crude oil, ES = E-mini S&P 500, GC = gold, NG = natural gas
    • FX (foreign exchange): Australian dollar, euro, British pound, Canadian dollar and Japanese yen against the US dollar

The Panel: Median Volatility and Variables

Slide 12 in the PDF

Three asset classes guard against a narrow result: a gain that appears in one class only does not show that the method transfers. The median volatility in the table is computed as \(\sqrt{252\,\mathrm{median}(\mathrm{RV}_t)}\) for each asset, then as the median over the class; 252 trading days a year turn a daily variance into an annual one, and the data contain trading days only.

The variables are five-minute RV, the semivariances \(\mathrm{RV}^+\) and \(\mathrm{RV}^-\), realised quarticity (RQ) and the daily open and close prices.

In brief

  • Three asset classes
    • a gain that appears in one class only does not show that the method transfers
  • Median volatility in the table: \(\sqrt{252\,\mathrm{median}(\RV_t)}\) per asset, then the median over the class
    • 252 trading days a year turn a daily variance into an annual one
    • The data contain trading days only
  • Variables
    • five-minute RV, semivariances RV\(^+\), RV\(^-\), realised quarticity (RQ)
    • daily open and close prices

Realised Volatility Across Asset Classes (1/2)

Slide 13 in the PDF

The chart shows annualised volatility \(\sqrt{252\,\mathrm{RV}_t}\) in percent, as a five-day mean on a log scale, for two assets per class: JPMorgan Chase (JPM) and Nvidia (NVDA) among stocks, the E-mini S&P 500 (ES) and crude oil (CL) among futures, and EURUSD (US dollars per euro) and USDJPY (yen per US dollar) among currencies.

    Question for the room We keep the same scale. When ▶ widens the moving-average window of the annualised volatility from 5 to 65 days, what happens to the crude-oil (CL) maximum in the first half of 2020?
    Answer

    It falls, from 458% (five days ending 22 April 2020) to 152% (65 days ending 3 June 2020). A longer averaging window mixes the extreme values with quieter days. The option “unchanged” confuses the smoothed series with the original observations; “rises” confuses a mean with a sum. With the five-day mean, ES reaches 115% in March, JPM 111% and NVDA 118%, while EURUSD and USDJPY stay near 20%. The median volatility per class is 17.1% for stocks, 19.7% for futures and 7.3% for FX.

    All classes share the spike of March 2020, while crude oil peaks in April 2020, so a forecaster must adapt within days. The continuous VOLARE series for CL joins successive contracts: the negative price of 20 April 2020 concerned only the May contract, close to expiry, and the series stays positive.

    In brief

    • Annualised volatility \(\sqrt{252\,\RV_t}\) in %, five-day mean, log scale, two assets per class
      • stocks: JPM = JPMorgan Chase, NVDA = Nvidia
      • futures: ES = E-mini S&P 500, CL = crude oil
      • FX: EURUSD = US dollars per euro, USDJPY = yen per US dollar
    • All classes share March 2020; crude oil (CL) peaks in April 2020
      • a forecaster must adapt within days
    • CL: the continuous VOLARE series joins successive contracts
      • the negative price of 20 April 2020 concerned only the May contract, close to expiry
      • the series stays positive

    Realised Volatility Across Asset Classes (2/2)

    Slide 14 in the PDF

    On a log scale volatility is persistent, the long memory that HAR approximates. Spikes decay within weeks, so the shape of the recent path may carry information beyond a few averages; this is the possibility that signature features are meant to exploit.

    In brief

    • On a log scale volatility is persistent (the long memory HAR approximates)
    • Spikes decay within weeks, so the shape of the recent path may carry information
    MFM_ch20_case_study

    Section summary

    The target is the average RV over 1, 5 or 22 days, measured with error that is largest on volatile days and ranked with the robust QLIKE loss. The panel covers 50 assets in three classes, with a common volatility spike in March 2020.

    Benchmarks: the HAR Family

    A new method is only interesting if it beats the models already in use. This section presents HAR, its log version and two extensions, and arranges the seven models of the chapter so that the contribution of each ingredient of the transferred method can be isolated.

    HAR: Three Averages of the Past (1/2)

    Slide 15 in the PDF

    The question is which simple benchmark must be beaten before a new method is believed. In the HAR model of Corsi (2009), future RV is a linear function of the daily, weekly and monthly averages of past RV:

    \[\mathrm{RV}_{t+1:t+h} = \beta_0 + \beta_d\,\mathrm{RV}_t + \beta_w\,\mathrm{RV}_t^{(w)} + \beta_m\,\mathrm{RV}_t^{(m)} + \varepsilon_{t+h},\]
    (20.7)

    with \(\mathrm{RV}_t^{(w)} = \frac{1}{5}\sum_{j=0}^{4}\mathrm{RV}_{t-j}\) and \(\mathrm{RV}_t^{(m)} = \frac{1}{22}\sum_{j=0}^{21}\mathrm{RV}_{t-j}\) the weekly and monthly averages up to day \(t\), and \(\varepsilon_{t+h}\) the forecast error.

    \(\beta_0\) is the intercept, and \(\beta_d\), \(\beta_w\) and \(\beta_m\), nonnegative in practice, measure how much yesterday, last week and last month weigh; they are estimated by OLS for each horizon \(h\). A sum \(\beta_d + \beta_w + \beta_m\) close to 1 means high persistence. The motivation is a cascade of traders at daily, weekly and monthly frequencies, which mimics long memory with three parameters.

    In brief

    • Question: which simple benchmark must be beaten before we believe a new method?
    • The HAR model Corsi (2009): future RV is a linear function of the daily, weekly and monthly averages of past RV
      \begin{equation*} \RV_{t+1:t+h} = \beta_0 + \beta_d\,\RV_t + \beta_w\,\RV_t^{(w)} + \beta_m\,\RV_t^{(m)} + \varepsilon_{t+h} \end{equation*}
      \begin{equation*} \RV_t^{(w)} = \frac{1}{5}\sum_{j=0}^{4}\RV_{t-j}, \qquad \RV_t^{(m)} = \frac{1}{22}\sum_{j=0}^{21}\RV_{t-j} \end{equation*}
    • Notation
      • \(\RV_t^{(w)}\), \(\RV_t^{(m)}\): weekly (5-day) and monthly (22-day) averages of RV up to day \(t\); \(\varepsilon_{t+h}\): forecast error
      • \(\beta_0\): intercept; \(\beta_d, \beta_w, \beta_m \ge 0\) in practice: how much yesterday, last week and last month weigh; estimated by OLS for each horizon \(h\)
      • \(\beta_d + \beta_w + \beta_m\) close to 1: high persistence of volatility
    • Traders at daily, weekly and monthly frequencies
      • a cascade that mimics long memory with three parameters

    HAR: Three Averages of the Past (2/2)

    Slide 16 in the PDF

    HAR is a restricted AR(22), an autoregression with 22 lags whose coefficients take only three distinct values: \(\phi_1 = \beta_d + \beta_w/5 + \beta_m/22\), \(\phi_{2..5} = \beta_w/5 + \beta_m/22\) and \(\phi_{6..22} = \beta_m/22\), where \(\phi_j\) is the coefficient of \(\mathrm{RV}_{t+1-j}\).

    The log-HAR model runs the same regression in logs,

    \[\ln \mathrm{RV}_{t+1:t+h} = \beta_0 + \beta_d \ln\mathrm{RV}_t + \beta_w \ln\mathrm{RV}^{(w)}_t + \beta_m \ln\mathrm{RV}^{(m)}_t + \varepsilon_{t+h},\]
    (20.8)

    which rules out negative forecasts and is less driven by spikes. Logs of averages are not averages of logs, so this log-HAR is not an exact restricted AR(22) in \(\ln \mathrm{RV}\) (Seminar 20, A7). The HARQ and SHAR (Semivariance HAR model) extensions are in the Appendix.

    In brief

    • HAR is a restricted AR(22): the 22 lag coefficients take only three distinct values
      \begin{equation*} \phi_1 = \beta_d + \frac{\beta_w}{5} + \frac{\beta_m}{22}, \qquad \phi_{2..5} = \frac{\beta_w}{5} + \frac{\beta_m}{22}, \qquad \phi_{6..22} = \frac{\beta_m}{22} \end{equation*}
      • \(\phi_j\): the coefficient of \(\RV_{t+1-j}\) in an AR(22) (autoregression with 22 lags)
    • log-HAR: the same regression in logs
      \begin{equation*} \ln\RV_{t+1:t+h} = \beta_0 + \beta_d \ln\RV_t + \beta_w \ln\RV^{(w)}_t + \beta_m \ln\RV^{(m)}_t + \varepsilon_{t+h} \end{equation*}
      • No negative forecasts, less driven by spikes
      • Logs of averages are not averages of logs, so this log-HAR is not an exact restricted AR(22) in \(\ln\RV\) (Seminar 20, A7)

    Seven Models in One Table

    Slide 17 in the PDF

    The table lists the seven models. HAR, HARQ and SHAR forecast RV in levels with equal weights: HARQ adds the term \(\mathrm{RV}_t(\mathrm{RQ}_t^{1/2} - \overline{\mathrm{RQ}^{1/2}})\) to HAR, and SHAR replaces \(\mathrm{RV}_t\) by the semivariances \(\mathrm{RV}_t^+\) and \(\mathrm{RV}_t^-\). The four remaining models forecast \(\ln \mathrm{RV}\):

    • log-HAR with equal weights;
    • log-HAR-K, the log-HAR regressors with signature-kernel weights;
    • Sig-L, the log-HAR regressors plus 36 non-constant signature terms (39 before dropping constants) selected by a two-step LASSO;
    • Sig-LK, Sig-L with kernel weights, which is the method of Gu et al. (2024).

    In the names, L stands for LASSO and K for kernel weights.

    log-HAR, log-HAR-K, Sig-L and Sig-LK form a \(2\times2\) ablation, signature features (yes/no) by kernel weights (yes/no), in which each ingredient is switched on and off separately. log-HAR is the natural baseline for the signature models, because it has the same target and its regressors are among the candidates. The LASSO may also drop log-HAR terms, however, so Sig-L against log-HAR compares two procedures, not only two feature sets.

    In brief

    ModelTargetWeightsRegressors
    HAR\(\RV\)equal\(\RV_t, \RV_t^{(w)}, \RV_t^{(m)}\)
    log-HAR\(\ln \RV\)equalthe same, in logs
    HARQ\(\RV\)equalHAR + \(\RV_t\,(\mathrm{RQ}_t^{1/2} - \overline{\mathrm{RQ}^{1/2}})\)
    SHAR\(\RV\)equal\(\RV_t^{+}, \RV_t^{-}, \RV_t^{(w)}, \RV_t^{(m)}\)
    log-HAR-K\(\ln \RV\)signature kernellog-HAR regressors
    Sig-L\(\ln \RV\)equallog-HAR + 36 non-constant signature terms (39 before dropping constants), two-step LASSO
    Sig-LK\(\ln \RV\)signature kernelas Sig-L: the method of Gu et al.
    • Names: L = LASSO, K = kernel weights
      • the charts call Sig-L “Signature LASSO”, Sig-LK “Signature LASSO + kernel weights”, log-HAR-K “log-HAR + kernel weights”
    • log-HAR, log-HAR-K, Sig-L and Sig-LK form a \(2 \times 2\) ablation
      • each ingredient switched on and off separately
      • signature features (yes/no) \(\times\) kernel weights (yes/no)
    • Log-HAR is the natural baseline for the signature models
      • same target, and its regressors are among the candidates
      • but the LASSO may also drop log-HAR terms
      • so Sig-L vs log-HAR compares two procedures, not only two feature sets

    Path Signatures

    The transferred method summarises a window of data by its path signature. This section defines the object: the window as a path, the signature as the sequence of iterated integrals, its computation for piecewise-linear paths through Chen's identity, the Lévy area as a measure of order, and the signature kernel that measures the similarity of two windows.

    A Window of Data Is a Path

    Slide 18 in the PDF

    The question is how to summarise a 22-day window so that the order of the moves is kept, not only their levels. The last \(L\) days of \(d-1\) series are taken, time is added as the first channel (time augmentation) and the points are connected linearly:

    \[X_{u_i} = \Big(u_i,\; \ln \mathrm{RV}_{t-L+1+i},\; \textstyle\sum_{j \le t-L+1+i} r_j\Big), \qquad u_i = \frac{i}{L-1},\; i = 0, \dots, L-1,\]
    (20.9)

    a path \(X: [0,1] \to \mathbb{R}^d\) with \(L = 22\) days and \(d = 3\) channels: time, log RV and the cumulative return. Here \(u_i \in [0,1]\) is the rescaled time of the \(i\)-th point and \(r_j = \ln(C_j/O_j)\) the open-to-close log return of day \(j\), with \(O_j\) and \(C_j\) the open and close prices.

    A regression on the window sees 22 lags; a path summary sees their order, speed and co-movement. It can distinguish, for example, a window in which RV rose while prices fell from one in which RV rose first and prices fell later.

    In brief

    • Question: how do we summarise a 22-day window so that the order of the moves is kept, not only their levels?
    • Take the last \(L\) days of \(d - 1\) series
      • add time as the first channel (time augmentation)
      • connect the points linearly
      \begin{equation*} X: [0, 1] \to \mathbb{R}^{d}, \qquad X_{u_i} = \Big(u_i,\; \ln \RV_{t-L+1+i},\; \textstyle\sum_{j \le t-L+1+i} r_j\Big), \qquad u_i = \frac{i}{L-1},\; i = 0, \dots, L - 1 \end{equation*}
      • \(L\): window length in days (here 22); \(d\): number of channels (here \(d = 3\): time, log RV, cumulative return); \(u_i \in [0, 1]\): rescaled time of the \(i\)-th point; \(r_j = \ln(C_j/O_j)\): open-to-close log return of day \(j\), with \(O_j\), \(C_j\) the open and close prices
    • A regression on the window sees 22 lags; a path summary sees their order, speed and co-movement
      • Example: RV rose while prices fell, or RV rose first and prices fell later

    Where Signatures Come From

    Slide 19 in the PDF

    Signatures come from the geometry of loop spaces. Chen (1957) showed that the iterated integrals of a path form an algebra, and Chen (1958) that they determine the path up to reparametrisation and retraced pieces. In the rough path theory of Lyons (1998), the signature is the natural feature set for controlled differential equations: a response to a path depends on it, to first order, only through its signature.

    Machine learning has used signatures as features in Levin, Lyons & Ni (2013) and Chevyrev & Kormilitzin (2016), and Lyons & McLeod (2022) survey the field; in finance, Cuchiero, Gazzani & Svaluto-Ferro (2023) build signature-based asset price models.

    In brief

    The Signature: Definition (1/2)

    Slide 20 in the PDF

    Level 1 of the signature keeps only the total increments of the channels. Integrals over ordered times keep, in addition, the order in which the channels move, which is why the signature is built from iterated integrals. One signature term is an iterated integral of the path along a word of channels:

    \[S(X)^{i_1 \dots i_k}_{a,b} = \int_{a < u_1 < \dots < u_k < b} dX^{i_1}_{u_1}\cdots dX^{i_k}_{u_k}.\]
    (20.10)

    Here \(X: [a,b] \to \mathbb{R}^d\) is a path of bounded variation (finite total length), for example piecewise linear, and \(X^i_u\) its channel \(i\) at time \(u\). The word \((i_1, \dots, i_k)\) is an ordered list of \(k\) channels with letters in \(\{1, \dots, d\}\), and \(k\) is the level of the term. The integral runs over ordered times \(u_1 < \dots < u_k\), and \(dX^i_u\) is the increment of channel \(i\) at time \(u\), for a linear segment its slope times \(du\).

    In brief

    • Why iterated integrals
      • level 1 keeps only the total increments
      • integrals over ordered times keep the order in which the channels move
    • One signature term is an iterated integral of the path along a word of channels
      \begin{equation*} S(X)^{i_1 \dots i_k}_{a,b} = \int_{a < u_1 < \dots < u_k < b} dX^{i_1}_{u_1}\cdots dX^{i_k}_{u_k} \end{equation*}
    • Notation
      • \(X: [a, b] \to \mathbb{R}^d\): a path of bounded variation (finite total length), e.g. piecewise linear; \(X^i_u\): its channel \(i\) at time \(u\)
      • Word \((i_1, \dots, i_k)\): an ordered list of \(k\) channels, letters in \(\{1, \dots, d\}\); \(k\) is the level of the term
      • \(u_1 < \dots < u_k\): the integral runs over ordered times; \(dX^{i}_{u}\): increment of channel \(i\) at time \(u\) (for a linear segment, slope times \(du\))

    The Signature: Definition (2/2)

    Slide 21 in the PDF

    The signature is the whole sequence of levels,

    \[\mathrm{Sig}(X) = \big(1,\; S^{(1)},\; S^{(2)},\; S^{(3)}, \dots\big), \qquad S^{(k)} \in (\mathbb{R}^d)^{\otimes k},\]
    (20.11)

    where \(S^{(k)}\) collects the \(d^k\) terms of level \(k\), one for each word of length \(k\), stored as a tensor (\(\otimes\) is the tensor product). The truncated signature \(\mathrm{Sig}^N(X)\) keeps levels 0 to \(N\); \(N\) is the depth.

    Level 1 contains the total increments, \(S^i = X^i_b - X^i_a\). Level 2 contains \(S^{ij} = \int_a^b (X^i_u - X^i_a)\,dX^j_u\), the increments of channel \(j\) weighted by how far channel \(i\) has already moved. Each further level adds one integral and records finer details of the order of the moves.

    In brief

    • The signature is the whole sequence of levels
      \begin{equation*} \Sig(X) = \big(1,\; S^{(1)},\; S^{(2)},\; S^{(3)}, \dots\big), \qquad S^{(k)} \in (\mathbb{R}^d)^{\otimes k} \end{equation*}
      • \(S^{(k)}\): the \(d^k\) terms of level \(k\), one for each word of length \(k\), stored as a \(k\)-dimensional array (a tensor; \(\otimes\): tensor product)
      • \(\Sig^N(X)\): the truncated signature, levels \(0\) to \(N\); \(N\) is the depth
    • Level 1: the total increments of the channels
      \begin{equation*} S^{i} = X^{i}_b - X^{i}_a \end{equation*}
    • Level 2: increments of channel \(j\) weighted by how far channel \(i\) has already moved
      \begin{equation*} S^{ij} = \int_a^b (X^{i}_u - X^{i}_a)\,dX^{j}_u \end{equation*}
      • Each level adds one integral and records finer details of the order of the moves

    Linear Segments and Chen’s Identity (1/2)

    Slide 22 in the PDF

    A linear segment with increment \(\Delta = X_b - X_a \in \mathbb{R}^d\) has the tensor exponential as signature, \(\mathrm{Sig} = \exp_\otimes(\Delta) = (1, \Delta, \Delta^{\otimes2}/2!, \Delta^{\otimes3}/3!, \dots)\), where \(\Delta^{\otimes k}\) is the array of all products \(\Delta^{i_1}\cdots\Delta^{i_k}\), so that a level-\(k\) term is \(\Delta^{i_1}\cdots\Delta^{i_k}/k!\).

    Chen's identity (Chen, 1957) states that the signature of two paths run one after the other is the tensor product of their signatures:

    \[\mathrm{Sig}(X * Y) = \mathrm{Sig}(X) \otimes \mathrm{Sig}(Y), \qquad (A \otimes B)^{(k)} = \sum_{p=0}^{k} A^{(k-p)} \otimes B^{(p)},\]
    (20.12)

    where \(X * Y\) is the concatenation, \(Y\) starting where \(X\) ends, and \(A^{(k)}\), \(B^{(k)}\) are level \(k\) of the signatures \(A\) and \(B\), with \(A^{(0)} = B^{(0)} = 1\). Level \(k\) of the whole path collects every way of splitting a word of length \(k\) between the first and the second piece.

    In brief

    • A linear segment with increment \(\Delta \in \mathbb{R}^d\) has the tensor exponential as signature
      \begin{equation*} \Sig(\text{segment}) = \exp_{\otimes}(\Delta) = \Big(1,\; \Delta,\; \tfrac{1}{2!}\Delta^{\otimes 2},\; \tfrac{1}{3!}\Delta^{\otimes 3}, \dots\Big) \end{equation*}
      • \(\Delta = X_b - X_a\): the increment of the segment; \(\Delta^{\otimes k}\): the array of all products \(\Delta^{i_1}\cdots\Delta^{i_k}\), so a level-\(k\) term is \(\Delta^{i_1}\cdots\Delta^{i_k}/k!\)
    • Chen’s identity Chen (1957)
      • the signature of two paths run one after the other is the tensor product of their signatures
      \begin{equation*} \Sig(X * Y) = \Sig(X) \otimes \Sig(Y), \qquad \big(A \otimes B\big)^{(k)} = \sum_{p=0}^{k} A^{(k-p)} \otimes B^{(p)} \end{equation*}
      • \(X * Y\): concatenation, \(Y\) starts where \(X\) ends; \(A^{(k)}\), \(B^{(k)}\): level \(k\) of the signatures \(A\), \(B\), with \(A^{(0)} = B^{(0)} = 1\)
      • Level \(k\) of the whole path collects every way of splitting a word of length \(k\) between the first and the second piece

    Linear Segments and Chen’s Identity (2/2)

    Slide 23 in the PDF

    For a piecewise-linear path with \(m\) segments and increments \(\Delta_1, \dots, \Delta_m\), (20.12) gives \(\mathrm{Sig} = \exp_\otimes(\Delta_1)\otimes\cdots\otimes\exp_\otimes(\Delta_m)\). This is the whole algorithm: \(m\) updates, each costing \(O(d^N)\) at depth \(N\).

    Rolling windows use Chen's identity in the other direction, to remove an initial piece: \(\mathrm{Sig}(X_{[s,t]}) = \mathrm{Sig}(X_{[a,s]})^{-1}\otimes\mathrm{Sig}(X_{[a,t]})\) for \(a \le s \le t\), where the inverse for \(\otimes\) is the signature of the piece run backwards.

    In brief

    • A piecewise-linear path with \(m\) segments, increments \(\Delta_1, \dots, \Delta_m\)
      \begin{equation*} \Sig = \exp_{\otimes}(\Delta_1) \otimes \cdots \otimes \exp_{\otimes}(\Delta_m) \end{equation*}
      • This is the whole algorithm: \(m\) updates, each costing \(O(d^N)\) at depth \(N\)
    • Rolling windows: Chen also removes an initial piece
      \begin{equation*} \Sig(X_{[s,t]}) = \Sig(X_{[a,s]})^{-1} \otimes \Sig(X_{[a,t]}) \end{equation*}
      • \(X_{[s,t]}\): the path restricted to \([s, t]\), with \(a \le s \le t\); \((\cdot)^{-1}\): inverse for \(\otimes\), the signature of the piece run backwards

    Worked Example: a Two-Segment Path

    Slide 24 in the PDF

    The worked example takes the path \((0,0) \to (1,0) \to (1,2)\), with increments \(\Delta_1 = (1,0)\) and \(\Delta_2 = (0,2)\). Level 1 is \(S^{(1)} = \Delta_1 + \Delta_2 = (1,2)\). Level 2 follows from Chen's identity, \(S^{ij} = \tfrac12\Delta_1^i\Delta_1^j + \Delta_1^i\Delta_2^j + \tfrac12\Delta_2^i\Delta_2^j\), with \(\Delta_m^i\) the component \(i\) of the increment of segment \(m\), and gives \(S^{11} = 1/2\), \(S^{12} = 2\), \(S^{21} = 0\) and \(S^{22} = 2\).

    Two checks follow:

    1. \(S^{11} = (\Delta X^1)^2/2\) and \(S^{22} = (\Delta X^2)^2/2\), an instance of the shuffle identity: the product of two level-1 coordinates equals the sum of their two orderings at level 2.
    2. \(S^{12} \neq S^{21}\), because channel 1 moves in the first segment and channel 2 in the second. Level 1 cannot tell this path from the one that goes up first and right second, since both have \(\Delta X = (1,2)\); level 2 tells them apart.

    In brief

    • The path \(X: (0,0) \to (1,0) \to (1,2)\), with increments \(\Delta_1 = (1, 0)\) and \(\Delta_2 = (0, 2)\)
    • Level 1: \(S^{(1)} = \Delta_1 + \Delta_2 = (1, 2)\)
    • Level 2 by Chen’s identity: \(S^{ij} = \tfrac12\Delta_1^i\Delta_1^j + \Delta_1^i\Delta_2^j + \tfrac12\Delta_2^i\Delta_2^j\)
      • \(\Delta_m^i\) is the component \(i\) of the increment of segment \(m\)
    • Result: \(S^{11} = \tfrac12\), \(\;S^{12} = 0 + 1\cdot 2 + 0 = 2\), \(\;S^{21} = 0\), \(\;S^{22} = \tfrac12\cdot 2^2 = 2\)
    • Check: \(S^{11} = (\Delta X^1)^2/2\) and \(S^{22} = (\Delta X^2)^2/2\)
      • shuffle identity: the product of two level-1 coordinates equals the sum of their two orderings at level 2
    • \(S^{12} \neq S^{21}\): the order matters, since channel 1 moves in the first segment and channel 2 in the second
    • The total displacement (level 1) cannot tell this path from the one that goes up first and right second
      • both have \(\Delta X = (1, 2)\); level 2 tells them apart

    The Lévy Area: Order as a Number

    Slide 25 in the PDF

    The Lévy area is the antisymmetric part of level 2,

    \[A^{ij} = \tfrac12\big(S^{ij} - S^{ji}\big),\]
    (20.13)

    equal to the signed area between the path, projected on channels \((i,j)\), and its chord, the straight line from start to end. \(A^{ij} > 0\) means that the path turns anticlockwise in the \((i,j)\) plane. The symmetric part, \(\tfrac12(S^{ij} + S^{ji}) = \tfrac12\Delta X^i\Delta X^j\), adds no new information. In the worked example, \(A^{12} = (2-0)/2 = 1\), the area of the triangle between the path and the chord.

    As a lead–lag statistic the area needs care. The lead–lag reading requires a positive association: if \(i\) falls and then \(j\) rises, \(i\) moves first, yet \(A^{ij} < 0\). For the channels (return, log RV), a price fall followed by a volatility rise gives \(A < 0\), and so does a price rise followed by a volatility fall. Leverage-type sequences (Chapter 5) are one contributor to a negative area, but the sign alone does not identify them.

    In brief

    • The Lévy area \(A^{ij}\): the antisymmetric part of level 2
      \begin{equation*} A^{ij} = \tfrac12\big(S^{ij} - S^{ji}\big) \end{equation*}
      • Equals the signed area between the path, projected on channels \((i, j)\), and its chord (the straight line from start to end)
      • \(A^{ij} > 0\): the path turns anticlockwise in the \((i, j)\) plane
      • The lead-lag reading (who moves first) needs a positive association: if \(i\) falls and then \(j\) rises, \(i\) moves first, yet \(A^{ij} < 0\)
    • Symmetric part: \(\tfrac12(S^{ij} + S^{ji}) = \tfrac12\,\Delta X^i\,\Delta X^j\), no new information
      • Worked example: \(A^{12} = (2 - 0)/2 = 1\), the area of the triangle between the path and the chord
    • Channels (return, log RV): a price fall followed by a volatility rise gives \(A < 0\)
      • so does a price rise followed by a volatility fall
    • A lead–lag statistic
      • leverage-type sequences (Chapter 5) are one contributor to a negative area
      • the sign alone does not identify them

    The Signature Kernel

    Slide 26 in the PDF

    The signature kernel is the inner product of the truncated signatures (Király & Oberhauser, 2019), a similarity score between two paths \(X\) and \(Y\) with the same channels:

    \[k_N(X,Y) = \langle \mathrm{Sig}^N(X), \mathrm{Sig}^N(Y)\rangle = 1 + \sum_{k=1}^{N}\sum_{|w|=k} S(X)^w S(Y)^w,\]
    (20.14)

    where \(w\) is a word, \(|w|\) its length and \(S(X)^w\) the term of word \(w\); the leading 1 is level 0. The distance induced by the kernel (Gu et al., Eq. 9) is the squared Euclidean distance between the truncated signatures, \(d_{\mathrm{Sig}}(X,Y) = k_N(X,X) - 2k_N(X,Y) + k_N(Y,Y) = \|\mathrm{Sig}^N(X) - \mathrm{Sig}^N(Y)\|^2\): zero when the two windows have the same signature up to depth \(N\), large when their shapes differ.

    Gu et al. (2024) and this chapter use this truncated kernel, with the same depth \(N\) as the features. The untruncated kernel solves a Goursat PDE (partial differential equation) (Salvi et al., 2021), which the paper cites but does not use. The channels are rescaled before the kernel is computed (Appendix); otherwise the channel with the largest units dominates.

    In brief

    • Kernel = inner product of the truncated signatures Király & Oberhauser (2019): a similarity score between two paths
      \begin{equation*} k_N(X, Y) = \langle \Sig^N(X), \Sig^N(Y) \rangle = 1 + \sum_{k=1}^{N}\sum_{|w| = k} S(X)^{w} S(Y)^{w} \end{equation*}
      • \(X\), \(Y\): two paths with the same channels; \(w\): a word, \(|w|\): its length (level); \(S(X)^{w}\): the term of word \(w\); the leading 1 is level 0
    • Distance induced by the kernel (Gu et al., Eq. 9)
      • squared Euclidean distance between the truncated signatures
      \begin{equation*} d_{\Sig}(X, Y) = k_N(X, X) - 2\,k_N(X, Y) + k_N(Y, Y) = \|\Sig^N(X) - \Sig^N(Y)\|^2 \end{equation*}
      • \(d_{\Sig} = 0\): the two windows have the same signature up to depth \(N\); large \(d_{\Sig}\): different shapes
    • Gu et al. and this chapter use this truncated kernel, with \(N\) as for the features
      • The untruncated kernel solves a Goursat PDE (partial differential equation) Salvi et al. (2021); the paper cites it but does not use it
    • Channels are rescaled before the kernel (Appendix): otherwise the channel with the largest units dominates

    Section summary

    A window of data becomes a path, and its signature records the increments and the order of the moves level by level. Chen's identity makes the signature of a piecewise-linear path a product of tensor exponentials, the Lévy area turns order into a number, and the signature kernel compares windows by the distance between their truncated signatures.

    The Paper: Gu et al. (2024)

    Before the method is transferred, the section states exactly what Gu et al. (2024) did and how strong their evidence is: the forecasting problem, the features, the two-step LASSO, the kernel weights, the reported gains and the limits of that evidence.

    The Problem at Amazon

    Slide 27 in the PDF

    The question is what exactly the paper did, and how strong its evidence is. Gu et al. (2024) forecast weekly North American truckload rates, nationally and for five regions, in 2018–2022, up to 12 weeks ahead. More than 100 candidate factors were screened down to about 40. The method has been used in Amazon operations since November 2022 and was published at KDD 2024 (Knowledge Discovery and Data Mining, the conference of the Association for Computing Machinery, ACM).

    The design is direct: one regression per horizon \(\Delta t\), in weeks, of the rate \(y_{\tau+\Delta t}\) on the features \(\xi_\tau\) observed at week \(\tau\).

    In brief

    • Question: what exactly did the paper do, and how strong is its evidence?
    • Weekly North American truckload rates, national and five regions, 2018–2022 Gu et al. (2024)
      • forecast up to 12 weeks ahead
      • More than 100 candidate factors screened to about 40
      • In Amazon operations since November 2022; KDD 2024 (Knowledge Discovery and Data Mining, the conference of the ACM, Association for Computing Machinery)
    • One direct regression per horizon \(\Delta t\) (in weeks; the paper's notation, not an increment \(\Delta\)), \(y_{\tau + \Delta t}\) on the features \(\xi_\tau\) (next slide)
    The Problem at Amazon A container truck on an American interstate, 2025 Photo: DanTD (2025); CC BY 4.0; Wikimedia Commons

    The Features of the Paper

    Slide 28 in the PDF

    The features at week \(\tau\) (Section 4.1, Eq. 14) combine external factors \(x_\tau\) with the signature of the time-augmented path of the target:

    \[\xi_\tau = \big[x_\tau,\; \mathrm{Sig}^N(\hat y_{\tau-l:\tau})\big], \qquad \hat y_{\tau-l:\tau} = \{(t_i, y_{t_i})\},\]
    (20.15)

    where \(y_{t_i}\) is the rate in week \(t_i\), \(\tau - l \le t_i \le \tau\), and \(l\) the lookback in weeks. The external factors bring the level and information from outside the series; the signature brings the shape of the recent path.

    In brief

    • Features at week \(\tau\) (Section 4.1, Eq. 14)
      • external factors \(x_\tau\)
      • the signature of the time-augmented target path
      \begin{equation*} \xi_\tau = \big[x_\tau,\; \Sig^N(\hat y_{\tau - l : \tau})\big], \qquad \hat y_{\tau-l:\tau} = \{(t_i, y_{t_i})\} \end{equation*}
      • \(y_{t_i}\): the rate in week \(t_i\), \(\tau - l \le t_i \le \tau\)
      • \(l\): lookback in weeks
      • \(\hat y_{\tau-l:\tau}\): the time-augmented path of the last \(l\) weeks
    • The external factors bring the level and information from outside the series
      • the signature brings the shape of the recent path

    The Method: Two-Step LASSO

    Slide 29 in the PDF

    The first step selects features with a weighted LASSO (Tibshirani, 1996) (Eq. 14):

    \[\hat\theta_{\mathrm{L}} = \arg\min_\theta \sum_\tau w_\tau\big(y_{\tau+\Delta t} - \xi_\tau^\top\theta\big)^2 + \lambda\|\theta\|_1,\]
    (20.16)

    with \(\theta\) the coefficient vector, \(\|\theta\|_1 = \sum_j|\theta_j|\), \(w_\tau \ge 0\) the weight of week \(\tau\), equal or given by the kernel, and \(\lambda \ge 0\) the penalty: a larger \(\lambda\) generally favours sparser solutions, and \(\lambda = 0\) gives weighted OLS. The second step (Eq. 15) refits by weighted OLS on the selected support, the nonzero coefficients of the first step.

    The refit is needed because the LASSO shrinks the kept coefficients towards zero; OLS after selection removes this penalty-induced shrinkage on the selected support (Belloni & Chernozhukov, 2013), but not selection or omitted-variable bias. The paper calls the method "adaptive" because of the weights \(w_\tau\), not in the sense of the adaptive LASSO of Zou (2006). In this chapter \(\lambda\) is chosen in each training window as the support with the smallest BIC (Bayesian Information Criterion) along the LARS (Least Angle Regression) path (Efron et al., 2004); details are in the Appendix.

    In brief

    • Step 1, selection by a weighted LASSO Tibshirani (1996) (Eq. 14)
      \begin{equation*} \hat\theta_{\mathrm{L}} = \arg\min_\theta \sum_{\tau} w_\tau\big(y_{\tau+\Delta t} - \xi_\tau^\top\theta\big)^2 + \lambda\|\theta\|_1 \end{equation*}
      • \(\theta\): coefficient vector, \(\|\theta\|_1 = \sum_j |\theta_j|\); \(w_\tau \ge 0\): weight of week \(\tau\) (equal, or from the kernel, next slide)
      • \(\lambda \ge 0\): penalty; a larger \(\lambda\) generally favours sparser solutions; \(\lambda = 0\) is weighted OLS
    • Step 2, refit by weighted OLS on the selected support only, the nonzero coefficients of step 1 (Eq. 15)
      \begin{equation*} \hat\theta = \arg\min_{\mathrm{supp}(\theta) = \mathrm{supp}(\hat\theta_{\mathrm{L}})} \sum_{\tau} w_\tau\big(y_{\tau+\Delta t} - \xi_\tau^\top\theta\big)^2 \end{equation*}
    • Why refit: LASSO shrinks the kept coefficients towards zero
      • OLS after selection removes the penalty-induced shrinkage on the selected support Belloni & Chernozhukov (2013)
      • the refit does not remove selection or omitted-variable bias
      • The paper calls it “adaptive” because of the weights \(w_\tau\), not in the sense of the adaptive LASSO of Zou (2006)
    • Choice of \(\lambda\) in this chapter: the support with the smallest BIC Efron et al. (2004)
      • BIC: Bayesian Information Criterion
      • along the LARS (Least Angle Regression) path
      • in each training window (Appendix)

    The Method: Kernel Weights

    Slide 30 in the PDF

    The kernel weights start from the distance of each past window to the last window with an observed target (Eq. 11), \(\delta_\tau = d_{\mathrm{Sig},N}(z_{\tau-l:\tau},\, z_{t-\Delta t-l:t-\Delta t})\). Here \(z_\tau = (x_\tau, y_{\tau+\Delta t})\) are the pairs of predictors and later target, not the features \(\xi_\tau\) (Algorithm 1), \(t\) is the current week and \(t - \Delta t\) the last week whose target \(y_t\) is already observed. The weights are a softmax with temperature \(\gamma \ge 0\) (Eq. 12):

    \[w_\tau = \frac{\exp(-\gamma\,\delta_\tau)}{\sum_s \exp(-\gamma\,\delta_s)},\]
    (20.17)

    where \(s\) runs over all training weeks, so the weights sum to 1. Similar windows, with small \(\delta_\tau\), get more weight; \(\gamma = 0\) gives ordinary equal-weight regression and \(\gamma \to \infty\) uses only the most similar past window.

    The idea is a nearest-neighbour scheme inside a regression: learn from the past periods that looked like today. Its cost is a smaller effective sample, measured by Kish's \(n_{\mathrm{eff}} = 1/\sum_\tau w_\tau^2\), which equals \(n\) for equal weights \(1/n\) and 1 if one week takes all the weight.

    In brief

    • Distance of each past window to the last window with an observed target (Eq. 11)
      \begin{equation*} \delta_\tau = d_{\Sig, N}\big(z_{\tau - l : \tau},\; z_{t - \Delta t - l : t - \Delta t}\big) \end{equation*}
      • \(z_\tau = (x_\tau, y_{\tau+\Delta t})\): the pairs of predictors and later target, not the features \(\xi_\tau\) (Algorithm 1); \(z_{\tau - l : \tau}\): their path over \(l\) weeks
      • \(t\): the current week; \(t - \Delta t\): the last week whose target \(y_t\) is already observed; \(d_{\Sig, N}\): the signature distance at depth \(N\)
    • Softmax weights with temperature \(\gamma \ge 0\) (Eq. 12): similar windows (small \(\delta_\tau\)) get more weight
      \begin{equation*} w_\tau = \frac{\exp(-\gamma\,\delta_\tau)}{\sum_{s} \exp(-\gamma\,\delta_s)} \end{equation*}
      • \(s\): runs over all training weeks, so the weights sum to 1; \(\gamma = 0\): ordinary (equal-weight) regression; \(\gamma \to \infty\): only the most similar past window
    • A nearest-neighbour idea inside a regression: learn from the past periods that looked like today
      • Cost of the idea: fewer effective observations, \(n_{\mathrm{eff}} = 1 / \sum_\tau w_\tau^2\) (Kish); \(n_{\mathrm{eff}} = n\) for equal weights \(1/n\), \(n_{\mathrm{eff}} = 1\) if one week takes all the weight

    What the Paper Reports

    Slide 31 in the PDF

    For national rates three months ahead, April–November 2021 (Table 1), the paper reports a relative error \(|y - \hat y|/|y|\), with \(y\) the realised rate and \(\hat y\) its forecast, of about 1–2%, against 13–27% for the industry forecast: "more than fivefold". Over horizons of 1–12 weeks (Table 2) the errors are 1.06–2.60% nationally and 1.05–5.69% by region, with an estimated annual saving of about 50 million dollars.

    The kernel ablation (Table 3) uses six weekly values, 31 October – 5 December 2021, for one region: errors reach 9.85% without the kernel and 5.15% with it. What the result supports is a comparison of procedures against a single commercial forecast, without standard errors.

    In brief

    • National rates, three months ahead, April–November 2021 (Table 1)
      • relative error \(|y - \hat y|/|y|\) (\(y\): realised rate, \(\hat y\): its forecast)
      • about 1–2% vs 13–27% for the industry forecast, “more than fivefold”
      • Horizons 1–12 weeks (Table 2): 1.06–2.60% nationally, 1.05–5.69% by region; estimated annual saving of about 50 million dollars
      • Kernel ablation (Table 3, six weekly values, 31 October–5 December 2021, one region): errors up to 9.85% without the kernel vs up to 5.15% with it
    • Type of evidence: comparison of procedures (it shows which method works better on the same data, not why)
      • against a single commercial forecast, without standard errors (next slide)

    What the Paper Reports: Limits of the Evidence

    Slide 32 in the PDF

    The evidence has three limits:

    1. There is one benchmark, an unnamed commercial forecast, with no ARIMA (autoregressive integrated moving average) model, no LASSO without signatures and no random walk.
    2. The evaluation is short: a few months and a few series, without standard errors or a test of equal accuracy.
    3. The tuning is undisclosed: the lookback \(l\), the depth \(N\), the temperature \(\gamma\), the penalty \(\lambda\) and how they were chosen.

    None of this makes the method wrong; it makes the evidence thin, which is why the method is tested elsewhere. The order is replication before extension, as in Chapter 19.

    In brief

    • One benchmark: an unnamed commercial forecast
      • no ARIMA (autoregressive integrated moving average) model
      • no LASSO without signatures, no random walk
    • Short evaluation: a few months, a few series, no standard errors, no test of equal accuracy
    • Undisclosed tuning: \(l\), \(N\), \(\gamma\), \(\lambda\) and how they were chosen
    • None of this makes the method wrong; it makes the evidence thin
      • which is why it is tested elsewhere
      • replication before extension (Chapter 19)

    What Transfers to Volatility: Kept and Adapted

    Slide 33 in the PDF

    Three elements of the method are kept:

    1. The signature of the time-augmented recent path of the target;
    2. The two-step LASSO with weights and a direct regression per horizon;
    3. Softmax weights \(\exp(-\gamma\,\delta_\tau)\) with the truncated signature kernel.

    The adaptations follow from the new target. RV is daily and noisy rather than a weekly rate, so the models work in logs, inference is HAC and the losses are robust. The external factors \(x_\tau\) are replaced by the log-HAR terms, which carry the level.

    The kernel is also adapted. It compares the paths (time, \(\ln\mathrm{RV}\), return) ending at \(\tau\) and at the origin \(t\), updated every 5 days, rather than the pairs \((x_\tau, y_{\tau+\Delta t})\) ending at \(t - \Delta t\). The choices the paper did not disclose, \(l\), \(N\), \(\gamma\) and \(\lambda\), are fixed in advance, and strong benchmarks with formal tests replace the single commercial forecast.

    In brief

    Transfers
    • Signature of the time-augmented recent path of the target
    • Two-step LASSO with weights, direct regression per horizon
    • Softmax weights \(\exp(-\gamma\,\delta_\tau)\) with the truncated signature kernel
    Must be adapted
    • Daily, noisy target instead of weekly rates: logs, HAC inference, robust losses
    • External factors \(x_\tau\) replaced by the log-HAR terms (they carry the level)
    • The kernel compares the paths (time, \(\ln\RV\), return) ending at \(\tau\) and at \(t\)
      • updated every 5 days
      • not the pairs \((x_\tau, y_{\tau+\Delta t})\) ending at \(t - \Delta t\)
    • Undisclosed choices (\(l\), \(N\), \(\gamma\), \(\lambda\)) fixed ex ante
      • strong benchmarks and formal tests replace the single commercial forecast

    The Transfer Design (Pre-Registered)

    A method with several tuning choices can always be made to look good on some asset, horizon or period. The design therefore fixes, before any result is seen, the hypotheses, the tuning rule for the kernel temperature, the rolling estimation scheme and the evaluation protocol, and it discloses the one change made after a test run.

    Four Hypotheses Fixed Before the Results

    Slide 34 in the PDF

    The question is what must be fixed before the results, so that a gain cannot be found by searching. Four hypotheses are stated:

    • H1 (features): signature terms add information beyond log-HAR, tested by Sig-L against log-HAR, with the same target and the same equal weights.
    • H2 (weights): kernel weights improve a given regression, tested by Sig-LK against Sig-L and log-HAR-K against log-HAR.
    • H3 (method): the full method of Gu et al. (2024) beats the HAR family, tested by Sig-LK against HAR and against log-HAR and by the model confidence set (MCS) of all seven models.
    • H4 (stress): any gain is larger in stress periods, the COVID-19 crash and the US tariff shock of 2–30 April 2025, where similar past windows should matter most.

    In brief

    • Question: what do we fix before the results, so that a gain cannot be found by searching?
    • H1 (features): signature terms add information beyond log-HAR
      • Test: Sig-L vs log-HAR, same target, same equal weights
    • H2 (weights): kernel weights improve a given regression
      • Tests: Sig-LK vs Sig-L and log-HAR-K vs log-HAR
    • H3 (method): the full method of Gu et al. beats the HAR family
      • Sig-LK vs HAR and vs log-HAR
      • model confidence set, MCS, of all seven models
    • H4 (stress): any gain is larger in stress periods
      • COVID-19 crash, 2–30 April 2025, the US tariff shock
      • where similar past windows should matter most

    Kernel Temperature: One Choice per Class, on a Validation Block

    Slide 35 in the PDF

    The kernel temperature is made scale-free and set in every training window as

    \[\gamma = \frac{c}{\mathrm{median}_\tau(\delta_\tau)},\]
    (20.18)

    where \(\delta_\tau\) are the kernel distances of the training days to the current window and \(c > 0\) is a dimensionless constant, the only tuning parameter. With \(c = 1\), a window at the median distance gets weight \(e^{-1}\) relative to an identical one; a larger \(c\) concentrates the weights.

    \(c\) is chosen from the grid \(\{0.5, 1, 2, 4\}\) per horizon and asset class, on the first 250 forecast days of each asset pooled within the class. The criterion is the class mean of QLIKE\((c)\) divided by the asset's average over the grid. Evaluation starts after the last validation target of the class, so no tuning target postdates an evaluated origin.

    For the 40 stocks the smallest temperature, \(c = 0.5\), wins at every horizon for Sig-LK, which gives the most nearly equal weights. For futures and FX the choice changes with the horizon, a sign of a flat, noisy criterion; all values are in the Appendix.

    In brief

    • Scale-free temperature, set in every training window
      \begin{equation*} \gamma = \frac{c}{\mathrm{median}_\tau(\delta_\tau)} \end{equation*}
      • \(\delta_\tau\): kernel distances of the training days to the current window; \(c > 0\): a dimensionless constant, the only tuning parameter
      • \(c = 1\): a window at the median distance gets weight \(e^{-1}\) relative to an identical one; larger \(c\): more concentrated weights
    • Grid \(c \in \{0.5, 1, 2, 4\}\), chosen per horizon and asset class
      • on the first 250 forecast days of each asset, pooled within the class
      • Criterion: mean over the class of QLIKE\((c)\) divided by the asset’s average over the grid
      • Evaluation starts after the last validation target of the class, so no tuning target postdates an evaluated origin
    • For the 40 stocks the smallest temperature, \(c = 0.5\), wins at every horizon for Sig-LK
      • the most nearly equal weights
      • for futures and FX the choice changes with the horizon
      • a flat, noisy criterion (all values in the Appendix)

    Rolling Estimation, Step by Step

    Slide 36 in the PDF

    The diagram shows the rolling scheme. The models are re-estimated every 5 trading days on the last 1,000 complete pairs \((\xi_\tau, y_{\tau+h})\) with \(\tau + h \le t\), so that every target used in estimation is observed at the origin \(t\). Between re-estimations the coefficients are fixed while the regressors are updated daily, and the forecasts from \(t\) to \(t+4\) use the same estimates. The kernel distances are recomputed at each re-estimation, against the window ending at the origin \(t\).

    Evaluation Protocol (1/3): the DM Test

    Slide 37 in the PDF

    The question is whether the loss difference between two models exceeds sampling noise. For each asset and horizon the mean QLIKE and MSE of two models are compared by the DM (Diebold–Mariano) test of equal predictive accuracy (Diebold & Mariano, 1995); the relative error of the paper is reported only descriptively (Appendix). The statistic is

    \[d_t = L(\mathrm{RV}_t, F^A_t) - L(\mathrm{RV}_t, F^B_t), \qquad \mathrm{DM} = \frac{\bar d}{\sqrt{\hat\Omega/T}},\]
    (20.19)

    where \(L\) is the loss (QLIKE or MSE), \(F^A_t\) and \(F^B_t\) the forecasts of models \(A\) and \(B\), \(\bar d\) the mean loss differential over the \(T\) evaluation days and \(\hat\Omega\) the long-run variance of \(d_t\) (Further Topics).

    Under \(H_0: \mathbb{E}[d_t] = 0\), DM is approximately \(N(0,1)\), and \(\mathrm{DM} < -1.645\) means that model \(A\) is significantly more accurate at 5%, one-sided. Because estimation is rolling, the test is read as the unconditional GW (Giacomini–White) test; the conditional GW test also asks whether \(d_{t-h}\) predicts \(d_t\), with instruments \((1, d_{t-h})\) (Giacomini & White, 2006).

    In brief

    • Question: is the loss difference between two models larger than sampling noise?
    • Per asset and horizon: mean QLIKE and MSE Diebold & Mariano (1995)
      • compared by the DM (Diebold–Mariano) test of equal predictive accuracy
      • the paper’s relative error is reported descriptively only (Appendix)
      \begin{equation*} d_t = L(\RV_t, F^{A}_t) - L(\RV_t, F^{B}_t), \qquad \mathrm{DM} = \frac{\bar d}{\sqrt{\hat\Omega / T}} \end{equation*}
      • \(L\): the loss (QLIKE or MSE); \(F^{A}_t\), \(F^{B}_t\): forecasts of models \(A\) and \(B\); \(d_t\): their loss differential
      • \(\bar d\): the mean of \(d_t\) over the \(T\) evaluation days; \(\hat\Omega\): the long-run variance of \(d_t\) (Further Topics)
    • Under \(H_0\colon \E[d_t] = 0\), DM is approximately \(N(0, 1)\)
      • \(\mathrm{DM} < -1.645\) means model \(A\) is significantly more accurate at 5% (one-sided)
    • Rolling estimation, so the test is read as the unconditional GW (Giacomini–White) test
      • The conditional GW test also asks whether \(d_{t-h}\) predicts \(d_t\) (instruments \((1, d_{t-h})\)) Giacomini & White (2006)

    Evaluation Protocol (2/3): the Model Confidence Set

    Slide 38 in the PDF

    The DM test compares two models, but here there are seven and no benchmark chosen in advance. The model confidence set of Hansen, Lunde & Nason (2011) at 90% is the set of models that contains the best one with probability at least 90%. It is built by elimination: while the \(T_{\max}\) statistic, the largest standardised gap between a model's mean loss and the average over the models still in the set, rejects equal accuracy, the worst model is dropped.

    The p-values come from the stationary bootstrap of Politis & Romano (1994), with 1,000 replications and a mean block length of \(\max(h, 10)\) days. The loss series of all models are resampled jointly, to keep the dependence across models.

    In brief

    • Why a set of models: the DM test compares two models
      • here there are seven and no benchmark chosen in advance
    • MCS at 90% Hansen, Lunde & Nason (2011): the set of models that contains the best one with probability at least 90%
      • Built by elimination: while the \(T_{\max}\) statistic (the largest standardised gap between a model’s mean loss and the average over the models still in the set) rejects equal accuracy, drop the worst model
      • \(p\)-values by the stationary bootstrap Politis & Romano (1994), 1000 replications, mean block length \(\max(h, 10)\) days
      • The loss series are resampled jointly, to keep the dependence across models

    Evaluation Protocol (3/3): Panel Test and Corrections

    Slide 39 in the PDF

    The panel test applies DM to the cross-sectional mean loss differential, one series with one value per date. In addition, the assets out of 50 with significantly better results are counted, with two corrections:

    • The Holm correction (Holm, 1979) controls the FWER (family-wise error rate), the probability of at least one false rejection;
    • The BH procedure of Benjamini & Hochberg (1995) controls the FDR (false discovery rate), the expected share of false rejections among the rejections.

    Results are also reported for subperiods: 20 February – 30 April 2020 (the COVID-19 crash), 2–30 April 2025 (the US tariff shock) and all other days. The evaluation periods by class are in the Appendix.

    In brief

    • Panel: DM test on the cross-sectional mean loss differential (one series, one value per date)
    • Count the assets (of 50) with significantly better results
      • The Holm correction Holm (1979) controls the FWER (family-wise error rate, the probability of at least one false rejection)
      • BH Benjamini & Hochberg (1995) controls the FDR (false discovery rate, the expected share of false rejections among the rejections)
    • Subperiods (evaluation periods by class: Appendix)
      • 20 Feb–30 Apr 2020 (COVID-19 crash)
      • 2–30 Apr 2025 (US tariff shock)
      • all other days

    A Disclosed Deviation

    Slide 40 in the PDF

    The pre-registration included the insanity filter of Bollerslev, Patton & Quaedvlieg (2016) (BPQ) on all models. A test run of the code on JPM, one asset and before any panel result, showed the filter replacing correct crash forecasts by the calm in-sample mean: on 12 March 2020 the realised next-day RV of JPM was about 40 times the filtered HAR forecast.

    The rule was therefore changed before the models were estimated on the 50-asset panel. The primary results impose positivity only: a forecast \(\le 0\) is replaced by the in-sample mean. The BPQ filter is kept as a robustness check and reported in full.

    In brief

    • Pre-registered: the BPQ (Bollerslev–Patton–Quaedvlieg) insanity filter on all models: a forecast outside the in-sample range of the target becomes the in-sample mean
      • A test run of the code on JPM (one asset, before any panel result) showed it replacing correct crash forecasts by the calm in-sample mean
      • On 12 March 2020 the realised next-day RV of JPM was about 40 times the filtered HAR forecast
    • Changed before estimating the models on the 50-asset panel
      • primary results with positivity only
      • a forecast \(\le 0\) becomes the in-sample mean
      • The BPQ filter is kept as a robustness check and reported in full

    Section summary

    The design fixes four hypotheses, one tuning constant per class chosen on a validation block, a rolling scheme with re-estimation every 5 days and an evaluation protocol of DM tests, model confidence sets and Holm and BH corrections. The single change to the pre-registered plan, the forecast filter, was made before any panel result and is disclosed.

    Results

    The section reports the pre-registered comparison on the 50 VOLARE assets: the QLIKE losses of every model relative to HAR, the asset-by-asset Diebold–Mariano (DM) tests with the Holm and Benjamini–Hochberg (BH) corrections, the model confidence sets, the panel tests, the COVID-19 crash and the kernel weights at the height of the crash. The main comparison is Sig-LK, the full method of Gu et al. (2024), against log-HAR, the simplest model in logs; HAR in levels is the benchmark that every ratio is expressed against.

    QLIKE Relative to HAR, Asset by Asset

    Slide 41 in the PDF
      Question for the room At h = 1, across all 50 assets, the mean QLIKE ratio of Sig-L to HAR is 0.886. We switch the benchmark from HAR to log-HAR. Where will the mean of the Sig-L ratios lie?
      Answer

      Above 1: the mean Sig-L/log-HAR ratio is 1.004, so Sig-L has a slightly higher relative loss in this sample; for Sig-LK, 1.045. When the benchmark changes, the page recomputes the ratio of each asset and then averages; it does not divide the two means. A value of 1 is guaranteed only for the benchmark compared with itself. Relative to HAR, the means are 0.883 for log-HAR, 0.885 for log-HAR-K and 0.886 for Sig-L. These means are not a significance test.

      The question is whether the method of Gu et al. (2024) beats the HAR models and, above all, log-HAR. For each asset the chart reports the ratio of the model's mean QLIKE loss to the mean QLIKE loss of HAR, so that a value below 1 means a more accurate forecast than HAR. The boxes give the interquartile range over the 50 assets, the line the median, the diamond the mean, and the whiskers extend to 1.5 interquartile ranges.

      The ratios are averaged over assets with equal weights, here as in the ablation and the filtered results. The mean of the ratios is not the ratio of the means, so a change of benchmark produces new per-asset ratios rather than a rescaling. Medians describe a typical asset, while means remain sensitive to a few large losses.

      In brief

      • Question: does the method of Gu et al. beat the HAR models and, above all, log-HAR?
      • Ratio for each asset: mean QLIKE of the model over mean QLIKE of HAR; below 1: more accurate than HAR
        • Boxes: interquartile range over the 50 assets; line: median; diamond: mean; whiskers: 1.5 interquartile ranges
      • Ratios are averaged over assets with equal weights
        • here, in the ablation and in the filtered results
        • the mean of the ratios is not the ratio of the means
        • so a new benchmark means new per-asset ratios
      • Medians describe a typical asset; means remain sensitive to a few large losses
      MFM_ch20_case_study

      Reading the QLIKE Comparison

      Slide 42 in the PDF

      At the one-day horizon, \(h = 1\), the mean ratios of log-HAR, log-HAR-K and Sig-L are close: 0.883, 0.885 and 0.886. The features alone, Sig-L against log-HAR, raise the QLIKE loss by 0.4%, so the 36 extra signature features bring no gain; the features together with the kernel weights, Sig-LK against log-HAR, raise it by 4.5%.

      At \(h = 5\) and \(h = 22\) days Sig-LK falls behind log-HAR, by 28.0% and 92.9%. Its median ratios to HAR, 0.997 and 0.968, are much better than its means, so a few assets and days drive the loss. Without the return channel in the path, Sig-LK reaches ratios of 0.878 and 0.770, close to those of log-HAR, 0.871 and 0.765.

      HARQ and SHAR lose to HAR in levels here, because the main design does not apply the BPQ (Bollerslev–Patton–Quaedvlieg) filter; the filter rescues HARQ (Appendix). The first verdict is that the method does not beat the simplest log model. A near-tie is not a proof of equal accuracy, and the result supports a comparison of procedures.

      In brief

      • \(h = 1\): log-HAR, log-HAR-K and Sig-L are close (0.883, 0.885, 0.886)
        • Features alone (Sig-L vs log-HAR): +0.4% in QLIKE, so 36 extra features bring no gain
        • Features and weights (Sig-LK vs log-HAR): +4.5%
      • \(h = 5, 22\): Sig-LK falls behind log-HAR (+28.0%, +92.9%)
        • Its medians relative to HAR (0.997, 0.968) are much better than its means: a few assets and days drive the loss
        • in a variant without the return channel, Sig-LK relative to HAR (0.878, 0.770) comes close to log-HAR (0.871, 0.765)
      • HARQ and SHAR (Semivariance HAR model) lose to HAR in levels here
        • because the main design does not apply the BPQ filter
        • the filter rescues HARQ (Appendix)
      • First verdict: the method does not beat the simplest log model
        • a near-tie is not a proof of equal accuracy
      • Type of evidence: comparison of procedures

      Asset by Asset: Wins and Losses, Corrected

      Slide 43 in the PDF

      The table counts, for each comparison and horizon, the assets on which the first model is significantly better and significantly worse under QLIKE, each as raw/Holm/BH counts. The test is the one-sided DM test at 5% of Diebold & Mariano (1995), and the family for the corrections is the 50 assets. The DM test needs finite moments of order \(2 + \delta\) and weak dependence of the QLIKE loss differential \(d_t\), not of the returns, and forecasts close to zero can violate these conditions.

      The two corrections control different error rates:

      • The procedure of Holm (1979) controls the family-wise error rate (FWER), the probability of at least one false rejection, under any dependence.
      • The procedure of Benjamini & Hochberg (1995) controls the false discovery rate (FDR), the expected share of false rejections among the rejections, under independence or under positive regression dependence on a subset (PRDS) (Benjamini & Yekutieli, 2001).

      In brief

      Comparison (QLIKE)\(h = 1\)\(h = 5\)\(h = 22\)
      Sig-LK vs log-HAR2/0/0  \(|\)  5/0/00/0/0  \(|\)  15/0/20/0/0  \(|\)  27/1/16
      Sig-L vs log-HAR3/1/1  \(|\)  5/0/00/0/0  \(|\)  9/0/01/0/0  \(|\)  21/0/0
      log-HAR-K vs log-HAR3/0/0  \(|\)  3/0/03/0/0  \(|\)  1/0/05/0/0  \(|\)  6/0/0
      Sig-LK vs Sig-L2/0/0  \(|\)  5/0/00/0/0  \(|\)  5/0/02/0/0  \(|\)  10/0/0
      Sig-LK vs HAR12/5/6  \(|\)  0/0/07/1/1  \(|\)  0/0/08/2/4  \(|\)  1/0/0
      log-HAR vs HAR21/5/17  \(|\)  0/0/023/13/20  \(|\)  0/0/042/25/42  \(|\)  0/0/0
      • Cell: significantly better on raw/Holm/BH assets \(|\) significantly worse on raw/Holm/BH assets
        • one-sided DM at 5%, family of 50 assets
        • the DM test needs finite moments of order greater than 2 and weak dependence of the QLIKE loss differential \(d_t\), not of the returns; forecasts close to zero can violate them
        • Holm controls the FWER under any dependence
        • BH controls the FDR under independence or PRDS (positive regression dependence on a subset) Benjamini & Yekutieli (2001)

      Asset by Asset: Reading the Table

      Slide 44 in the PDF

      For H1, the hypothesis that the signature features add information, Sig-L beats log-HAR on 3 of the 50 assets at \(h = 1\) before correction and on 1 after Holm, while about 2.5 raw wins are expected by chance alone. At the same horizon Sig-L loses to log-HAR on 5 assets before correction and on none after Holm.

      For H2, the hypothesis that the kernel weights help, neither log-HAR-K against log-HAR nor Sig-LK against Sig-L shows a Holm-significant difference at any horizon. Against HAR at \(h = 1\), Sig-LK wins on 12 assets, but log-HAR alone wins on 21. The benchmark must therefore isolate the idea being tested: a comparison with HAR credits the signature with a gain that belongs to the log transform.

      In brief

      • H1 (signatures): Sig-L beats log-HAR on 3 of 50 assets at \(h = 1\) raw, 1 after Holm
        • about 2.5 raw wins are expected by chance
      • Also at \(h = 1\), Sig-L loses to log-HAR on 5 assets raw and on none after Holm
      • H2 (weights): log-HAR-K vs log-HAR and Sig-LK vs Sig-L: no Holm-significant difference at any horizon
      • Against HAR at \(h = 1\), Sig-LK wins on 12 assets, but log-HAR alone wins on 21
        • choose the benchmark that isolates the idea

      Model Confidence Sets (1/2)

      Slide 45 in the PDF
        Question for the room We keep QLIKE and all 50 assets. When ▶ raises the MCS confidence level from 90% to 99%, what happens to the share of assets for which Sig-LK stays in the set?
        Answer

        It rises or stays the same. The elimination threshold α falls from 0.10 to 0.01, and the page keeps the models with an MCS p-value above the threshold, so it eliminates fewer. For Sig-LK, the shares go from 100%, 94% and 88% to 100%, 100% and 100% at h = 1, 5 and 22; for log-HAR, from 100%, 98% and 98% to 100%. A higher confidence level does not mean a more restrictive selection. Membership of the MCS does not establish equal accuracy: at 90%, the four log models are rarely separated, while HAR, HARQ and SHAR are excluded for a fifth to a third of the assets.

        The model confidence set (MCS) of Hansen, Lunde & Nason (2011) is the set of models that cannot be rejected as the best at a given level. The chart shows, for each model, the share of the 50 assets in which it belongs to the 90% MCS under QLIKE at \(h = 1\), 5 and 22. log-HAR belongs to it for 100%, 98% and 98% of the assets, and Sig-LK for 100%, 94% and 88%. The MCS needs the same moment and dependence conditions on the loss differentials between models as the DM test.

        The MCS rarely separates the four log models. HAR, HARQ and SHAR are excluded for a fifth to a third of the assets, so the level target is the clearest loser.

        In brief

        • Share of the 50 assets in which each model belongs to the 90% MCS under QLIKE (\(h = 1, 5, 22\))
          • log-HAR 100%, 98%, 98%
          • Sig-LK 100%, 94%, 88%
          • the MCS on QLIKE needs the same conditions as the DM test on the loss differentials between models, not on the returns
        • The MCS rarely separates the four log models
          • HAR, HARQ and SHAR are excluded for a fifth to a third of the assets
          • the level target is the clearest loser

        Model Confidence Sets (2/2)

        Slide 46 in the PDF

        Under the mean squared error (MSE) almost every model stays in the set. Differences in mean MSE loss have large standard errors, because the squared errors of variance forecasts are dominated by a few extreme days.

        In brief

        • Under MSE almost every model stays in the set
          • mean-loss differences have large standard errors
        MFM_ch20_case_study

        Panel Tests: Pooling the Evidence

        Slide 47 in the PDF

        The panel test pools the evidence: for each day the series is the mean QLIKE differential over the assets available that day, an unbalanced panel, and the DM statistic is computed on this series, with a negative \(t\) meaning that the first model is better. Pooled, only the log transform is clearly significant: log-HAR beats HAR with \(t = -3.09\), \(-3.95\) and \(-4.50\) at the three horizons, all with \(p < 0.001\).

        Sig-L against log-HAR at \(h = 1\) gives \(t = -1.91\), significant in a single one-sided test (\(p = 0.028\)). The result is fragile:

        • For the 40 stocks alone the sign flips, to \(t = +1.72\) against Sig-L;
        • After a Holm correction over the three horizons, with threshold \(0.05/3 \approx 0.017\), it is no longer significant;
        • It disappears in the balanced panel of the next slide.

        The conditional test of Giacomini & White (2006) (GW), Sig-LK against log-HAR, gives 0, 0 and 0 Holm rejections out of 50 assets.

        In brief

        Comparison: DM \(t\) (one-sided \(p\))\(h = 1\)\(h = 5\)\(h = 22\)
        Sig-LK vs log-HAR+1.45 (0.927)+1.68 (0.953)+1.80 (0.964)
        Sig-L vs log-HAR-1.91 (0.028)+1.48 (0.931)+1.51 (0.935)
        log-HAR-K vs log-HAR+1.89 (0.971)+0.40 (0.654)+1.92 (0.973)
        Sig-LK vs Sig-L+1.74 (0.959)+0.98 (0.837)-0.77 (0.221)
        log-HAR vs HAR-3.09 (\(<\)0.001)-3.95 (\(<\)0.001)-4.50 (\(<\)0.001)
        • Series: the mean QLIKE differential over the assets available that day (unbalanced panel)
          • negative \(t\) = first model better
          • the DM test needs finite moments of order greater than 2 and weak dependence of the QLIKE loss differential \(d_t\), not of the returns; forecasts close to zero can violate them
        • Pooled, only the log transform is clearly significant
        • Sig-L vs log-HAR at \(h = 1\): significant in a single one-sided test (\(p = 0.028\)) in the unbalanced panel
          • in the balanced panel the result disappears (next slide)
          • For the 40 stocks alone the sign flips: \(t = +1.72\) against Sig-L
        • After a Holm correction over the three horizons (threshold \(0.05/3 \approx 0.017\)) it is no longer significant
        • Conditional GW test, Sig-LK vs log-HAR: 0, 0 and 0 Holm rejections out of 50

        The Balanced Panel: the Same Assets Every Day

        Slide 48 in the PDF

        The unbalanced panel averages the assets available each day: futures and foreign exchange (FX) from 2014, the stocks only from 27 January 2020. Differences between asset classes are thus mixed with differences between periods. The balanced panel keeps all 50 assets but only the days on which all of them have forecasts, 1,511 days at \(h = 1\), from 27 January 2020 to 29 January 2026.

        On the balanced panel Sig-L against log-HAR at \(h = 1\) gives \(t = +0.75\) (\(p = 0.773\)), against \(t = -1.91\) (\(p = 0.028\)) in the unbalanced panel; on the common calendar only the panel of the 5 futures favours Sig-L (\(t = -4.03\)). The advantage of log-HAR over HAR remains, with \(t\) between −4.76 and −2.21.

        In brief

        Comparison: DM \(t\) (one-sided \(p\))\(h = 1\)\(h = 5\)\(h = 22\)
        Sig-LK vs log-HAR+1.19 (0.884)+1.61 (0.946)+1.41 (0.921)
        Sig-L vs log-HAR+0.75 (0.773)+1.49 (0.932)+1.24 (0.892)
        log-HAR vs HAR\(-\)2.21 (0.014)\(-\)2.51 (0.006)\(-\)4.76 (\(<\)0.001)
        • The unbalanced panel averages the assets available each day: futures and FX from 2014, the stocks only from 27 Jan 2020
          • differences between classes are thus mixed with differences between periods
        • The balanced panel: all 50 assets, only the days on which all of them have forecasts (1511 days at \(h = 1\), 27 Jan 2020 – 29 Jan 2026)
        • Sig-L vs log-HAR at \(h = 1\): \(t = +0.75\) (\(p = 0.773\)), against \(t = -1.91\) (\(p = 0.028\)) in the unbalanced panel
          • on the common calendar only the panel of the 5 futures favours Sig-L (\(t = -4.03\))
        • The advantage of log-HAR over HAR remains: \(t\) between \(-\)4.76 and \(-\)2.21
        MFM_ch20_case_study

        Inside the COVID-19 Crash: JPM, Next Day (1/2)

        Slide 49 in the PDF
          Question for the room At h = 1, which of the three displayed forecasts captures the size of the target jump for the 11 March 2020 origin of JPM volatility?
          Answer

          None. At this origin, the next-day target is about 139%, while the forecasts are 59% (HAR), 49% (log-HAR) and 60% (Sig-LK): all strongly underestimate the jump. The somewhat faster later reaction of Sig-LK in mid-March and the later overshooting of the target by HAR do not show that either anticipated this jump. The raw HAR forecast turns negative in late March; the chart shows the value after the positivity rule (the mean target of the training window), and the 7 replaced origins are marked with ×.

          The chart shows, for JPMorgan Chase (JPM) from February to May 2020, the one-day forecasts of the models and the target, both as annualised volatility in percent, \(\sqrt{252F_t}\) and \(\sqrt{252\,\mathrm{RV}_{t+1}}\). All models lag the first jump. Sig-LK reacts slightly faster than log-HAR in mid-March, when the path has already turned, and after the peak both decay at similar speeds.

          In brief

          • Annualised volatility in % of the forecasts, \(\sqrt{252 F_t}\), and of the target, \(\sqrt{252\,\RV_{t+1}}\)
            • February–May 2020
            • all models lag the first jump
          • Sig-LK reacts slightly faster than log-HAR in mid-March
            • when the path has already turned
            • after the peak both decay at similar speeds

          Inside the COVID-19 Crash: JPM, Next Day (2/2)

          Slide 50 in the PDF

          HAR in levels overshoots, and in late March its raw forecast turns negative. The chart shows the value after the positivity rule, the mean target of the training window, which appears as a flat segment; the 7 replaced origins, 27 March–6 April 2020, are marked with a cross. A model in levels can produce impossible variance forecasts exactly in a crisis, which the log target avoids by construction.

          In brief

          • HAR in levels overshoots, then the raw HAR forecast turns negative in late March
            • the chart shows the value after the positivity rule: the mean target of the training window (the flat segment)
            • the 7 replaced origins, 27 March – 6 April 2020, are marked with \(\times\)
          MFM_ch20_case_study

          Stress Periods vs Normal Times (H4) (1/2)

          Slide 51 in the PDF
            Question for the room At h = 1, in the COVID-19 crash, the mean QLIKE ratio Sig-LK/HAR is 0.73. Is this sufficient to credit the signature procedure with a gain over log-HAR?
            Answer

            No. The value 0.73 shows an advantage over HAR; with the same benchmark, however, log-HAR has 0.70, so the results show no additional advantage of Sig-LK. On the other days, the ratios are 0.94 (log-HAR) and 0.97 (Sig-LK). At h = 22, in the same crash, Sig-LK has 6.35 times and Sig-L 8.30 times the HAR loss. The results give no descriptive support to H4; they are not a formal test of the difference between periods.

            The chart compares, within subperiods fixed before the evaluation, the mean over assets of each model's QLIKE loss relative to HAR: the COVID-19 crash, the US tariff shock of April 2025 and the other days. The values are on the next slide.

            Stress Periods vs Normal Times (H4) (2/2)

            Slide 52 in the PDF

            H4 states that the signature models gain most in stress periods, when similar past windows should help. At \(h = 1\) in the COVID-19 crash the ratios to HAR are 0.70 for log-HAR, 0.73 for Sig-LK and 1.89 for HARQ; in the US tariff shock of 2–30 April 2025 they are 0.88 for log-HAR and 0.92 for Sig-LK, and on the other days 0.94 and 0.97.

            At \(h = 22\) in the COVID-19 crash the losses of Sig-LK and Sig-L are 6.35 and 8.30 times that of HAR. Exactly where similar past windows should help, the signature models fail most, so H4 is not supported. The result supports a description, since no formal test of the interaction between model and period was run.

            In brief

            • Mean over assets of QLIKE relative to HAR within each subperiod fixed before the evaluation
            • \(h = 1\), COVID-19 crash: log-HAR 0.70, Sig-LK 0.73, HARQ 1.89
              • 2–30 April 2025 (US tariff shock): 0.88 and 0.92
              • other days: 0.94 and 0.97
            • \(h = 22\), COVID-19 crash: Sig-LK 6.35 and Sig-L 8.30 times the loss of HAR
              • exactly where similar past windows should help, the signature models fail most
              • H4 not supported
            • Type of evidence: description (it shows what happened in the data, without testing a cause)
              • no formal interaction test
            MFM_ch20_case_study

            Kernel Weights at the Origin 16 March 2020 (1/2)

            Slide 53 in the PDF
              Question for the room At the 16 March 2020 origin, with N = 3, ▶ raises c from 0.5 to 8. What happens to the mean effective sample size over the 40 stocks?
              Answer

              It falls, from about 975 to 36 of 1,000. The weights are proportional to exp(−c δ / median of δ), where δ is the distance between windows: a larger c concentrates them on the small distances. The number of observations stays 1,000; the effective size falls. With c = 0.5, the average weight stays between 0.61 and 1.43 times 1/n; with c = 8, between 0.10 and 157 times 1/n. The highest weights go to the last days before the origin, whose 22-point windows overlap the current window. Translation invariance does not mean equal distances between windows.

              The chart shows the weights that Sig-LK (\(h = 1\)) gives at the origin of 16 March 2020 to the training days, 24 March 2016–13 March 2020, multiplied by the window length so that 1 means equal weighting. The weights are nearly equal, between 0.61 and 1.43 times \(1/n\), and the effective sample size is 975 of 1,000 observations on average over the 40 stocks.

              In brief

              • Weights of Sig-LK (\(h = 1\)) at the origin 16 March 2020 over the training days
                • training days: 24 Mar 2016 – 13 Mar 2020
                • weights times the window length: 1 = equal weighting
              • Nearly equal: between 0.61 and 1.43 times \(1/n\)
                • effective sample size 975 of 1000 (average over the 40 stocks)

              Kernel Weights at the Origin 16 March 2020 (2/2)

              Slide 54 in the PDF

              The highest weights go to the last days before the origin. Recent 22-day windows overlap the current window, so their similarity may partly reflect the overlap rather than a distinct economic regime. The late-2018 sell-off receives less weight, not more, and the weights are almost uncorrelated with log volatility (correlation +0.04).

              The reason is the translation invariance of the signature: it describes the shape of the path, not its level. A past crisis at a different volatility level does not look similar to the current one, so the kernel cannot implement "learn from past crises".

              In brief

              • The highest weights go to the last days before the origin
              • Recent 22-point windows overlap the current window
                • similarity may partly reflect overlap, not a distinct economic regime
              • The late-2018 sell-off gets less weight
                • weights are almost uncorrelated with log volatility (+0.04)
                • the translation-invariant signature sees shape, not level
              MFM_ch20_case_study

              Section summary

              Sig-LK does not beat log-HAR: at one day the two are close and at a week and a month the signature model loses clearly. The pooled tests confirm only the gain of the log transform over HAR, the apparent Sig-L gain at one day does not survive the balanced panel or the correction over horizons, and in the COVID-19 crash the kernel weights stay almost flat.

              Ablations and Robustness

              A null result for the full method leaves open which of its two ingredients failed. The ablation crosses the features (signature or not) with the weights (kernel or equal), and the results by asset class show where any gain is concentrated.

              The \(2 \times 2\) Ablation

              Slide 55 in the PDF
                Question for the room At h = 1, the mean QLIKE ratio Sig-LK/log-HAR over the 50 assets is 1.045, and the largest ratio is that of crude oil (CL). If we leave CL out, does the mean fall below 1?
                Answer

                No: the mean falls from 1.045 to 1.011, so leaving CL out does not reverse the descriptive ranking. CL has a ratio of 2.71, but the effect of leaving it out also depends on the other 49 ratios. The weights alone (log-HAR-K) give 1.002, 1.002 and 1.004 relative to log-HAR at h = 1, 5 and 22; Sig-L gives 1.004, 1.285 and 2.238; Sig-LK, 1.045, 1.280 and 1.929. Sig-L evaluates the procedure with signatures and LASSO selection, so its difference from log-HAR cannot be attributed to the features alone.

                The question is which ingredient drives the result, the signature features or the kernel weights. The chart reports the mean over the 50 assets of each model's QLIKE loss relative to log-HAR, for the four combinations of features (signature or log-HAR terms only) and weights (kernel or equal). The dashed line marks 1, the loss of log-HAR, so a bar above it means a worse forecast than log-HAR.

                In brief

                • Question: which ingredient drives the result, the signature features or the kernel weights?
                • Mean over the 50 assets of QLIKE relative to log-HAR
                  • features (signature or not) crossed with weights (kernel or equal)
                • The dashed line marks 1 (log-HAR)
                  • a bar above it means a worse forecast than log-HAR
                MFM_ch20_case_study

                The \(2 \times 2\) Ablation: the Numbers

                Slide 56 in the PDF

                The weights alone, log-HAR-K, give ratios of 1.002, 1.002 and 1.004 at \(h = 1\), 5 and 22: no detectable effect on the mean loss. The features alone, Sig-L, give 1.004, 1.285 and 2.238, and both together, Sig-LK, 1.045, 1.280 and 1.929. The average effective sample size with the Sig-LK weights is 816 of 1,000 observations.

                Adding the weights to Sig-L reduces the loss at \(h = 22\) without removing it. At \(h = 1\) they make it worse, mainly through crude oil (CL), with a ratio of 2.71, unlike the kernel gain reported in Table 3 of Gu et al. (2024). The loss of the signature models at longer horizons therefore comes from the features, not from the weights.

                In brief

                • Weights alone (log-HAR-K): 1.002, 1.002, 1.004 at \(h = 1, 5, 22\): no detectable effect on the mean loss
                • Features alone (Sig-L): 1.004, 1.285, 2.238: level-3 terms explode in crashes (Further Topics)
                • Both (Sig-LK): 1.045, 1.280, 1.929
                  • Average effective sample size with the Sig-LK weights: 816 of 1000
                  • Adding the weights to Sig-L reduces the loss at \(h = 22\) without removing it
                  • They worsen \(h = 1\) through crude oil (CL: 2.71), unlike the kernel gain in Table 3 of Gu et al.

                By Asset Class

                Slide 57 in the PDF

                By asset class, the ratio of Sig-L to log-HAR is 1.009, 1.332 and 2.505 for the 40 stocks at \(h = 1\), 5 and 22, 0.965, 1.084 and 1.206 for the 5 futures, and 0.996, 1.111 and 1.134 for the 5 FX pairs. The only gain is for futures at \(h = 1\), 3.5% in QLIKE, and marginally for FX.

                Five futures at one horizon are a candidate for a new, pre-registered study, not a finding of this one. The stocks, 80% of the panel, show no gain, and the mean of the ratios over assets is dominated by them. The DM test on the unbalanced panel, by contrast, mixes classes with periods, because before 2020 it holds only futures and FX.

                In brief

                Sig-L relative to log-HAR (QLIKE)\(h = 1\)\(h = 5\)\(h = 22\)
                Stocks (40)1.0091.3322.505
                Futures (5)0.9651.0841.206
                FX (5)0.9961.1111.134
                • The only gain: futures at \(h = 1\) (-3.5%) and, marginally, FX
                  • Five futures and one horizon: a candidate for a new, pre-registered study, not a finding of this one
                • Stocks, 80% of the panel, show no gain
                  • the mean of the ratios over assets is dominated by the stocks
                  • the DM test on the unbalanced panel, however, mixes classes with periods: before 2020 it holds only futures and FX

                Section summary

                The kernel weights have no detectable effect on the mean loss, and the signature features cause the losses at the week and month horizons. The one gain, for futures at one day, is too narrow to count as a result of this study.

                Limitations and Open Directions

                Limitations of This Study

                Slide 58 in the PDF

                The study evaluates one design out of many, so its null result is not a null result for signatures in general. The design fixes a window of \(L = 22\) days, depth \(N = 3\), three channels, selection by the Bayesian information criterion (BIC) and a four-point grid for the kernel temperature \(c\). Only the truncated kernel is used, and the external factors that are central in Gu et al. (2024), such as implied volatility or macroeconomic news, are absent.

                Two benchmark families lie outside the comparison:

                1. Realized GARCH (Hansen, Huang & Shek, 2012), where GARCH denotes generalised autoregressive conditional heteroskedasticity (Chapter 5);
                2. Machine-learning benchmarks (Chapter 13).

                The stocks are evaluated over six years only.

                In brief

                • One design out of many: a null result for this design is not one for signatures in general
                  • the design: \(L = 22\), \(N = 3\), three channels, BIC, a four-point grid for \(c\)
                • Truncated kernel only
                • No external factors (implied volatility, macro news), central in Gu et al.
                • Realized GARCH Hansen, Huang & Shek (2012) and machine-learning benchmarks are outside the comparison
                  • GARCH: generalised autoregressive conditional heteroskedasticity (Chapter 5)
                  • machine-learning benchmarks: Chapter 13
                • Stocks are evaluated over six years only

                Open Directions

                Slide 59 in the PDF

                Several directions remain open:

                • A level-aware kernel would add basepoint augmentation or the level of \(\ln\mathrm{RV}\) to the distance (Seminar 20, C2);
                • The path could be built from intraday 5-minute returns;
                • The signature kernel could be computed exactly by the partial differential equation (PDE) method of Salvi et al. (2021);
                • Log-signatures or bounded (tanh) channels could tame the explosive level-3 terms (Morrill et al., 2020).

                Cross-asset paths are another direction, since their Lévy areas measure lead–lag across markets, the subject of the case study below. Finally, the futures gain at \(h = 1\) could be tested on a longer panel with this design fixed in advance.

                In brief

                • A level-aware kernel: basepoint augmentation or the level of \(\ln\RV\) in the distance (Seminar 20, C2)
                • Signatures of the intraday path (5-minute returns)
                • The PDE kernel Salvi et al. (2021)
                • Log-signatures and bounded (tanh) channels against explosive level-3 terms Morrill et al. (2020)
                • Cross-asset paths, where Lévy areas measure lead–lag across markets
                • Futures at \(h = 1\) on a longer panel, with this design fixed in advance

                Case Study: The Signature as a Lead-Lag Metric

                The case study uses the signature for a different question: not to forecast one series, but to measure whether one asset moves before another. The Lévy area, a level-2 signature term, is a natural candidate, and the study illustrates it on intraday SPY and Bitcoin data.

                The Question: The Signature as a Lead-Lag Metric

                Slide 60 in the PDF

                The question is whether one asset moves before another and whether the signature of their joint path can measure it. Lead–lag relations show in which market information appears first, a question of price discovery. Bennett, Cucuringu and Reinert (2022, Machine Learning) build a directed network from pairwise lead–lag relations and split it into clusters of leaders and followers, using the 434 New York Stock Exchange (NYSE) stocks with the highest volume and daily returns from 2000 to 2019.

                One of the metrics they compare is twice the Lévy area of the path of normalised log prices over \([t-m, t]\):

                \[S^{ij} = 2A^{ij} = \iint_{t-m<u<v<t} \big(dX^i_u\, dX^j_v - dX^j_u\, dX^i_v\big).\]
                (20.20)

                A positive \(S^{ij}\) means that moves of \(i\) are followed by moves of \(j\) in the same direction. Unlike cross-correlations, (20.20) needs no choice of a maximum lag, but it assumes a positive association: it cannot tell "\(i\) leads with negative association" from "\(j\) leads with positive association".

                In brief

                • Question: does one asset move before another, and can the signature of their joint path measure it?
                  • lead-lag relations show in which market information appears first (price discovery)
                • Bennett, Cucuringu & Reinert (2022), Machine Learning
                  • the pairwise lead-lag relations form a directed network, which is then split into clusters of leaders and followers
                  • data: the 434 NYSE (New York Stock Exchange) stocks with the highest volume, daily returns 2000–2019
                • One of the metrics compared: twice the Lévy area of the path of normalised log prices over \([t - m, t]\)
                  \begin{equation*} S^{ij} = 2A^{ij} = \iint_{t-m<u<v<t} \big(dX^i_u\, dX^j_v - dX^j_u\, dX^i_v\big) \end{equation*}
                  • \(S^{ij} > 0\): moves of \(i\) are followed by moves of \(j\) in the same direction
                  • unlike cross-correlations, it needs no choice of a maximum lag
                  • it assumes a positive association: it cannot tell “\(i\) leads, with negative association” from “\(j\) leads, with positive association”

                What the Paper Found, and with Which Procedure

                Slide 61 in the PDF

                The empirical results of the paper come from a metric other than the signature (Sections 5–6 of the published version): the area under the cross-correlation function with distance correlation, with lags of up to 5 days, and the Hermitian random-walk (RW) clustering of the network. The lead–lag clusters are significant in a permutation test that shuffles the order of the days; that permutation destroys both the lagged dependence between assets and the autocorrelation of each asset, so on its own it does not separate them.

                In the authors' comparisons the clusters do not match Standard Industrial Classification (SIC) sectors, and their correlations with volume and market capitalisation are small. In Section 6.3, the signal built on the clusters has an annualised Sharpe ratio of 0.62, against 0.40 for the S&P 500 over the same period, both before transaction costs. The next slides illustrate the Lévy area on intraday data; they do not replicate the paper's procedure.

                In brief

                • The empirical results of the paper come from a metric other than the signature (published version, Sections 5–6)
                  • the metric: ccf-auc (area under the cross-correlation function) with distance correlation, with lags of up to 5 days
                  • the network clustering: Hermitian RW (random walk)
                • The lead-lag clusters are significant in a permutation test that shuffles the order of the days
                  • the permutation destroys both the lagged dependence between assets and the autocorrelation of each asset: on its own, the test does not separate them
                • In the authors' comparisons the clusters do not match SIC (Standard Industrial Classification) sectors
                  • their correlations with volume and with market capitalisation are small
                • The paper's numbers (Section 6.3): the signal built on these clusters has an annualised Sharpe ratio of 0.62
                  • the S&P 500 has 0.40 over the same period
                  • both are before transaction costs
                • The next slides: an illustration of the Lévy area on intraday data, not the paper's procedure

                SPY and Bitcoin at Five Minutes: Who Moves First?

                Slide 62 in the PDF
                  Question for the room Across all 489 days, what is the longest tested window for which the mean Lévy area between SPY and Bitcoin is positive and significant at 5% after the Holm correction for the 20 tests?
                  Answer

                  120 minutes: NW (Newey–West, standard errors) t ≈ 3.05 and Holm-adjusted p ≈ 0.032. At 195 minutes, NW t ≈ 2.45 passes the pointwise 1.96 threshold, but the adjusted p is 0.157. The 10-minute option confuses the window with the strongest signal (t = 9.6, the area positive on 70% of the days) with the longest significant window; 195 minutes uses the pointwise threshold instead of the correction. Over the whole session, t = 1.51. The correlation between SPY at t − 1 and BTC (Bitcoin) at t is 0.088 (NW t = 9.8), and in the reverse direction −0.005. The 20 tests (12 lags and 8 windows) are exploratory, and the result indicates a lead-lag association, not causality. The effect weakens in 2026: the lag +1 correlation falls to 0.04.

                  The data are five-minute bars of SPY, the S&P 500 exchange-traded fund (ETF), and of Bitcoin (BTC, in USD), New York time, 09:30–16:00, on the 489 days with all 78 bars for both assets out of 513 common days, 3 September 2024–18 September 2026. The left panel shows the mean over days of the correlation between the SPY return at \(t-l\) and the BTC return at \(t\), with pointwise 95% intervals from the Newey–West (NW) standard errors. The right panel shows the NW \(t\)-statistic of the mean daily Lévy area of (SPY, BTC) for windows from 10 minutes to the whole session; the line at 1.96 is a pointwise threshold.

                  In brief

                  • SPY (the S&P 500 ETF; ETF: Exchange-Traded Fund) and Bitcoin (BTC, in USD): five-minute bars, New York time, 09:30–16:00
                    • 489 days with all 78 bars for both assets (out of 513 common days), 3 Sep 2024 – 18 Sep 2026
                  • Left: mean over days of the correlation between the SPY return at \(t - l\) and the BTC return at \(t\); bars: pointwise 95% intervals with the Newey-West (NW) standard error of the tests
                  • Right: NW \(t\)-statistic of the mean daily Lévy area (SPY, BTC), for windows from 10 minutes to the whole session; the line at 1.96 is a pointwise threshold
                  MFM_ch20_leadlag

                  The Method in Brief and How the Replication Differs

                  Slide 63 in the PDF

                  The path of each day is formed by the cumulative sums of the SPY and BTC returns, each return divided by the standard deviation of its asset on that day, as the authors recommend, since otherwise the area grows with volatility. The 78 returns run from the open of the 09:30 bar to the close of each bar from 09:30 to 15:55. The Lévy area of a window of \(m\) steps is a sum of lagged products of the SPY returns \(x_i\) and the BTC returns \(y_i\):

                  \[A_m = \tfrac12 \sum_{l=1}^{m-1} \sum_{i} \big(x_i\, y_{i+l} - y_i\, x_{i+l}\big).\]
                  (20.21)

                  In (20.21) lag \(l\) appears \(m-l\) times, so a window of 2 steps, 10 minutes, keeps lag 1 only, while a long window adds many lags without effect and hence noise. The daily statistic is the mean area over all windows of \(m\) steps in the day, and the test is the NW \(t\) with 8 lags on the 489 daily values (Newey & West, 1987). The illustration differs from the paper: two assets instead of 434, five-minute bars instead of daily returns, and no network clustering or trading signal.

                  In brief

                  • The path of the day: cumulative sums of the SPY and BTC returns, each return divided by the standard deviation of its asset on that day
                    • 78 returns: from the open of the 09:30 bar to the close of each bar 09:30–15:55
                    • the normalisation the authors recommend: otherwise the area grows with volatility
                  • The Lévy area of a window of \(m\) steps is a sum of lagged products of the returns \(x_i\) (SPY) and \(y_i\) (BTC)
                    \begin{equation*} A_m = \tfrac12 \sum_{l=1}^{m-1} \sum_{i} \big(x_i\, y_{i+l} - y_i\, x_{i+l}\big) \end{equation*}
                    • lag \(l\) appears \(m - l\) times: a window of 2 steps (10 minutes) keeps lag 1 only
                    • a long window adds many lags without effect, hence noise
                  • Daily statistic: the mean area over all windows of \(m\) steps in the day
                    • the test: NW \(t\) with 8 lags on the 489 daily values
                  • Differences from the paper
                    • two assets instead of 434, and five-minute bars instead of daily returns
                    • no network clustering and no trading signal

                  The Result: SPY Leads Bitcoin by Five Minutes; Sensitivity to the Window Length

                  Slide 64 in the PDF

                  SPY moves before Bitcoin by one five-minute step. The correlation at lag +1 is 0.088 (NW \(t\) = 9.8), and at lag −1 it is −0.005 (\(t\) = −0.71); the direct test of direction, the difference between the two, is 0.093 (NW \(t\) = 9.5). The contemporaneous correlation is 0.40, and at lags +2 to +6 the correlations are small and negative, between −0.016 and −0.004.

                  The signature metric finds the effect only in short windows. With 10-minute windows \(t\) = 9.6 and the area is positive on 70% of the days; with 2-hour windows \(t\) = 3.0; over the whole session of 390 minutes \(t\) = 1.51 and the area is positive on 51% of the days. The analysis is exploratory, with 20 tests (12 non-zero lags and 8 windows): the Holm-adjusted p-value is 0.032 for the 2-hour window and 0.157 for 195 minutes. The effect weakens in 2026: the lag +1 correlation is 0.11 in 2024, 0.12 in 2025 and 0.04 in January–September 2026.

                  In brief

                  • SPY moves before Bitcoin by one five-minute step
                    • the correlation at lag \(+1\) is 0.088 (NW \(t\) = 9.8), and at lag \(-1\) it is \(-\)0.005 (\(t\) = \(-\)0.71)
                    • the direct test of direction: the difference between the correlations at lags \(+1\) and \(-1\) is 0.093 (NW \(t\) = 9.5)
                    • the contemporaneous correlation is 0.40; at lags \(+2\) to \(+6\) the correlations are small and negative (between \(-\)0.016 and \(-\)0.004)
                  • The signature metric finds the effect only in short windows
                    • 10 minutes: \(t\) = 9.6, and the area is positive on 70% of the days
                    • 2 hours: \(t\) = 3.0; the whole session (390 minutes): \(t\) = 1.51, area positive on 51% of the days
                  • The analysis is exploratory: 20 tests (12 non-zero lags and 8 windows)
                    • Holm-adjusted \(p\): 0.032 for the 2-hour window, 0.157 for 195 minutes
                  • The effect weakens in 2026: the lag \(+1\) correlation is 0.11 in 2024, 0.12 in 2025 and 0.04 in 2026 (January–September)

                  What the Lead-Lag Result Supports, and Its Limits

                  Slide 65 in the PDF

                  The result supports a five-minute lead–lag association between two assets, not a causal relation and not a profitable signal, since there are no trading costs and no out-of-sample evaluation. The sign of the area measures the orientation of the path, and reading it as "SPY leads" assumes a positive association between the two assets.

                  Three limits remain:

                  1. A one-step lag also arises from unsynchronised timestamps of the two data sources, and separating the two explanations needs tick data.
                  2. The window length \(m\) plays the role of the maximum lag, so the signature does not remove this choice.
                  3. The NW \(t\)-test needs finite moments and weak dependence of the daily statistic tested, bounded correlations and normalised areas, not of the raw returns.

                  In brief

                  • Type of evidence: a five-minute lead-lag association between two assets, not a causal relation and not a profitable signal
                    • no trading costs and no out-of-sample evaluation
                  • The sign of the area measures the orientation of the path; reading it as “SPY leads” assumes a positive association between the two assets
                  • Limits
                    • a one-step lag also arises from unsynchronised timestamps of the two sources; separating the two needs tick data
                    • the window length \(m\) plays the role of the maximum lag: the signature does not remove this choice
                    • the NW \(t\)-test needs finite moments and weak dependence of the daily statistic tested (bounded correlations, normalised areas), not of the raw returns

                  Section summary

                  The Lévy area detects the five-minute lead of SPY over Bitcoin only when the window is short, because longer windows add lags that carry no signal. The window length acts as the maximum lag of a correlation analysis, and the association is not a trading result.

                  AI for Scientific Discovery

                  The chapter is itself a transfer study: a method from a data-mining conference paper applied to a financial forecasting problem. The section turns this into a general question, states the criterion for a successful transfer and shows how the AI-assisted (Artificial Intelligence) loop was applied and where it can mislead.

                  When Does a Method Transfer Across Domains?

                  Slide 66 in the PDF

                  Methods travel faster than evidence: a Knowledge Discovery and Data Mining (KDD) conference paper on freight rates becomes a candidate for volatility, credit or energy. Two questions decide whether a transfer can work:

                  1. Which properties of the source problem made the method work;
                  2. Whether the target shares them.

                  Finance adds a hard filter: in Menkveld et al. (2024), 164 teams working on the same data disagree more than their standard errors suggest.

                  The criterion is that a transfer succeeds if the gain over the strongest simple benchmark survives across a pre-registered set \(\mathcal{S}\) of designs, assets and horizons, with the family corrected. In this chapter \(\mathcal{S}\) is 50 assets by 3 horizons by 4 variants. Holm and BH were applied per comparison and horizon, to families of 50 assets, and the variants are reported descriptively, without a joint correction over all 600 cells.

                  In brief

                  • Methods travel faster than evidence
                    • a KDD paper on freight rates becomes a candidate for volatility, credit, energy
                    • Which properties of the source problem made the method work?
                    • Does the target share them?
                  • Finance adds a hard filter Menkveld et al. (2024)
                    • 164 teams on the same data disagree more than their standard errors suggest
                  • Criterion: a transfer succeeds if the gain over the strongest simple benchmark survives
                    • a pre-registered set \(\mathcal{S}\) of designs, assets and horizons
                    • with the family corrected
                    • In this chapter: \(\mathcal{S}\) = 50 assets \(\times\) 3 horizons \(\times\) 4 variants; Holm/BH were applied per comparison and horizon (families of 50 assets), variants reported descriptively; no joint correction over all 600 cells

                  The Discovery Loop Applied to This Chapter

                  Slide 67 in the PDF

                  The loop has six steps:

                  1. In the reading step an assistant lists every equation, hyperparameter and benchmark of the paper and marks what is not disclosed, and each item is checked against the PDF.
                  2. The mapping step decides, for each component, whether it transfers, what must change and which benchmark isolates it.
                  3. The pre-registration fixes the hypotheses, grid, tests and correction in a dated document before the first result; the assistant drafts, the team decides every number.
                  4. The pieces are built and tested: signatures are checked against identities, such as level 1 equal to the increment and Chen's identity on a split path, not against another library.
                  5. The grid is run and then attacked, with Holm over 50 assets and a search for the single asset that would make a misleading headline.
                  6. The null is explained with data (checklist in the Appendix): here, flat weights caused by translation invariance and explosive level-3 terms caused by the exponential link.

                  In brief

                  • Read: an assistant lists every equation, hyperparameter and benchmark of the paper
                    • it marks what is not disclosed
                    • each item is checked against the PDF (prompts)
                  • Map: for each component, does it transfer, what must change, which benchmark isolates it
                  • Pre-register: hypotheses, grid, tests and correction
                    • fixed in a dated document before the first result
                    • the assistant drafts, the team decides every number
                  • Build and test the pieces: signatures against identities, not against another library
                    • level 1 = increment, Chen on a split path
                  • Run the grid, then attack it
                    • Holm over 50 assets
                    • a search for the single asset that would make a misleading headline (prompts)
                  • Explain the null, checked with data (checklist: Appendix)
                    • here flat weights (translation invariance)
                    • explosive level-3 terms (exponential link)

                  Mini-Case: Three Readings of the Same Results

                  Slide 68 in the PDF

                  Three headlines can be written from the same runs:

                  1. "Signatures beat HAR on 12 of 50 assets" (one-day horizon, raw DM) is true, but log-HAR alone beats HAR on 21, so the gain belongs to the log transform.
                  2. "Signatures improve E-mini S&P 500 volatility forecasts" (ES, \(h = 1\), DM \(t = -2.91\) against log-HAR) rests on one cell out of 150 asset-horizon cells; after Holm, 0 of 50 assets are significant at \(h = 1\) and none at \(h = 5\) or 22.
                  3. The headline of the pre-registered evaluation: "no robust advantage over log-HAR in the specification evaluated; worse at longer horizons; no detectable effect of the kernel weights".

                  An AI assistant asked to "find where the method works" will produce the first two headlines; the plan protects the third. The result supports a comparison of procedures: the same runs read without correction, on a single cell, or by the pre-registered plan.

                  In brief

                  • “Signatures beat HAR on 12 of 50 assets” (Sig-LK, one-day horizon, raw DM)
                    • True, but log-HAR alone beats HAR on 21: the gain belongs to the log transform
                  • “Signatures improve E-mini S&P 500 volatility forecasts” (ES, \(h = 1\), DM \(t = -2.91\) for Sig-LK against log-HAR)
                    • One cell out of 150 asset-horizon cells: after Holm, Sig-LK wins on 0 of 50 at \(h = 1\) and none at \(h = 5, 22\)
                  • The headline of the pre-registered evaluation
                    • “no robust advantage over log-HAR in the specification evaluated; worse at longer horizons; no detectable effect of the kernel weights”
                    • An AI (Artificial Intelligence) assistant asked to “find where the method works” will produce the first two; the plan protects the third
                  • Type of evidence: comparison of procedures
                    • the same runs, read without correction, on a single cell or by the pre-registered plan

                  Your Turn: Design a Transfer Study

                  Slide 69 in the PDF

                  The project picks one open direction of the chapter, such as a level-aware kernel, intraday signatures, log-signatures or futures only, and writes \(H_0\) and \(H_1\), the benchmark that isolates the idea and the result that would lead to dropping it. An assistant lists what could go wrong in the design, and each item gets one unit test.

                  The pre-registration is fixed in a dated document before the first out-of-sample number; then the grid is run and the distribution of results over \(\mathcal{S}\) is reported. Every AI error caught goes into `AI_ERRORS.md`; Seminar 20, C1, gives typical ones.

                  In brief

                  • Pick one open direction of this chapter
                    • level-aware kernel, intraday signatures, log-signatures, futures only
                    • Write \(H_0\)/\(H_1\), the benchmark that isolates it, and the result that would make you drop it
                  • Use an assistant to list what could go wrong in your design, then write one unit test per item
                  • Fix the pre-registration in a dated document before the first out-of-sample number
                    • run the grid
                    • report the distribution over \(\mathcal{S}\)
                  • Log every AI error you catch in AI_ERRORS.md (Seminar 20, C1 gives typical ones)

                  Conclusions

                  Key Takeaways

                  Slide 70 in the PDF

                  Signatures turn a window into a feature vector with a theory behind it: Chen's identity, uniqueness and universality. Transferred to realised variance, the method of Gu et al. (2024) does not beat log-HAR:

                  • H1, that signatures add information, is not supported, with no gain at one day and a loss at a week and a month.
                  • H2, that kernel weights help, is not supported: the weights are almost flat and down-weight past stress.
                  • H3 holds only against HAR in levels, through the log target.
                  • H4, gains in stress, is not supported.

                  Two mechanisms explain the null: the translation-invariant kernel sees shape, not level, and the level-3 terms explode in crashes. The strongest effect in the study is the oldest one, modelling the log of RV. The null result is informative, because the ablations and the mechanism explain why the method does not transfer.

                  In brief

                  • Signatures turn a window into a feature vector with a theory behind it
                    • Chen’s identity; uniqueness and universality (details in Further Topics)
                  • Transferred to realised variance, the method of Gu et al. does not beat log-HAR
                    • H1 (signatures add information): not supported, no gain at one day, a loss at a week and a month
                    • H2 (kernel weights help): not supported, weights are almost flat and down-weight past stress; H3: only HAR in levels is beaten, through the log target; H4 (gains in stress): not supported
                  • The translation-invariant kernel sees shape, not level
                    • level-3 terms explode in crashes
                  • The strongest effect in the study is the oldest one: model the log of RV
                  • The null result is informative
                    • the ablations and the mechanism explain why the method does not transfer

                  Readings on Signatures and Realised Measures

                  Slide 71 in the PDF

                  Introductions to signatures are Chevyrev & Kormilitzin (2016) and Lyons & McLeod (2022); realised measures are covered in Franke, Härdle & Hafner (2019) and Chapter 9. The VOLARE data are free for research at volare.unime.it.

                  Check your understanding

                  Slide 72 in the PDF

                  The questions check the chapter's main points:

                  1. What order information a signature keeps that an average loses;
                  2. Why both a level and a log target are compared;
                  3. Which ablation isolates the signatures;
                  4. Whether one significant win among many assets establishes superiority;
                  5. What the correction for multiple comparisons does;
                  6. How a null result can still be informative.

                  In brief

                  The Practical Decision

                  Slide 73 in the PDF

                  The chapter chooses log-HAR to forecast daily realised variance; the method of Gu et al. (2024), Sig-LK, does not replace it. log-HAR is in the 90% MCS for 100%, 98% and 98% of the assets at \(h = 1\), 5 and 22, and on the panel it beats HAR at every horizon, with DM \(t\) from −3.09 to −4.50.

                  The choice holds under three conditions:

                  1. QLIKE as the main loss, with the log target and the variance correction \(\hat s^2/2\);
                  2. Rolling estimation, re-estimated every 5 days on the last 1,000 pairs;
                  3. A panel dominated by stocks, 40 of the 50 assets.

                  It would be dropped if a gain of Sig-L or Sig-LK over log-HAR survived the Holm correction in a pre-registered study, for example the Sig-L gain on futures at \(h = 1\), −3.5% in QLIKE, confirmed on a longer panel.

                  In brief

                  • What I choose: log-HAR to forecast daily realised variance; the method of Gu et al. (Sig-LK) does not replace it
                    • log-HAR is in the 90% MCS for 100%, 98% and 98% of the assets at \(h = 1\), 5 and 22
                    • on the panel, log-HAR beats HAR at every horizon (DM \(t\) from \(-3.09\) to \(-4.50\))
                  • Under which conditions
                    • QLIKE as the main loss, with the log target and the variance correction \(\hat s^2/2\)
                    • rolling estimation, re-estimated every 5 days on the last 1000 pairs
                    • a panel dominated by stocks (40 of the 50 assets)
                  • What result would make me drop it
                    • a gain of Sig-L or Sig-LK over log-HAR that survives the Holm correction in a pre-registered study
                    • for example, the Sig-L gain on futures at \(h = 1\) (\(-3.5\%\) in QLIKE), confirmed on a longer panel

                  Further Topics

                  The Further Topics give the data background, the two refinements of HAR, the theory of signatures and details of the evaluation. They are reading for familiarity.

                  Why Five-Minute RV, and Where Do the Data Come From?

                  Slide 74 in the PDF

                  Alternatives to five-minute realised variance (RV) include bipower variation, robust to jumps, the realised kernel, robust to noise (Barndorff-Nielsen et al., 2008), and range-based estimators, but across 400 assets little beats five-minute RV as a proxy of daily variance (Liu, Patton & Sheppard, 2015). Its semivariances and quarticity are computed on the same grid, so SHAR and HARQ use consistent inputs, and every model is evaluated on the same target. VOLARE is built at the Universities of Messina and Florence from tick data with the same estimators for every asset class (Cipollini et al., 2026); it is free for research at volare.unime.it and is not redistributed with the course.

                  In brief

                  • Alternatives to RV
                  • Target of this chapter: five-minute RV
                  • Its semivariances and quarticity are on the same grid
                    • so SHAR (the semivariance HAR model) and HARQ use consistent inputs
                  • Every model is evaluated on the same target
                    • so loss differences come from the models, not from the volatility measure
                  • VOLARE is built at the Universities of Messina and Florence from tick data Cipollini et al. (2026)
                    • with the same estimators for every asset class
                  • Free for research at volare.unime.it, not redistributed with this course
                    • download the realised-variance files
                    • point the notebooks to your local folder
                  • The authors request the citation
                    • Cipollini, Cruciani, Gallo, Insana, Otranto & Spagnolo (2026), arXiv:2602.19732

                  Two Refinements (1/2): HARQ

                  Slide 75 in the PDF

                  RV measures the integrated variance with error, \(\mathrm{RV}_t = \mathrm{IV}_t + \eta_t\) with \(\mathrm{Var}(\eta_t) \approx 2\,\mathrm{IQ}_t/M\), and errors in variables pull the daily coefficient \(\beta_d\) towards zero, most on noisy days. HARQ (Bollerslev, Patton & Quaedvlieg, 2016) lets the weight on yesterday's RV depend on the square root of realised quarticity:

                  \[\mathrm{RV}_{t+1:t+h} = \beta_0 + \big(\beta_d + \beta_Q(\mathrm{RQ}_t^{1/2} - \overline{\mathrm{RQ}^{1/2}})\big)\mathrm{RV}_t + \beta_w\mathrm{RV}_t^{(w)} + \beta_m\mathrm{RV}_t^{(m)} + \varepsilon_{t+h},\]
                  (20.22)

                  with \(\beta_Q < 0\) expected, so that a noisy day gets less weight. The BPQ "insanity filter" replaces a forecast outside the in-sample range of the target by the in-sample mean.

                  In brief

                  • RV measures the true variance with error
                    \begin{equation*} \RV_t = \mathrm{IV}_t + \eta_t, \qquad \Var(\eta_t) \approx 2\,\mathrm{IQ}_t / M \end{equation*}
                    • \(\eta_t\): measurement error; errors-in-variables attenuate (pull towards zero) \(\beta_d\), most on noisy days
                  • HARQ Bollerslev, Patton & Quaedvlieg (2016): the weight on yesterday’s RV falls when RV is measured less precisely
                    \begin{equation*} \RV_{t+1:t+h} = \beta_0 + \big(\beta_d + \beta_Q(\mathrm{RQ}_t^{1/2} - \overline{\mathrm{RQ}^{1/2}})\big)\RV_t + \beta_w\RV_t^{(w)} + \beta_m\RV_t^{(m)} + \varepsilon_{t+h} \end{equation*}
                    • \(\mathrm{RQ}_t^{1/2}\): square root of realised quarticity, a measure of the error of \(\RV_t\); \(\overline{\mathrm{RQ}^{1/2}}\): its in-sample mean, so \(\beta_d\) is the weight on an average day
                    • Expected \(\beta_Q < 0\): on a noisy day (\(\mathrm{RQ}_t\) above average) yesterday’s RV gets less weight
                  • BPQ (Bollerslev–Patton–Quaedvlieg) filter (“insanity filter”)
                    • a forecast outside the in-sample range of the target is replaced by the in-sample mean

                  Two Refinements (2/2): SHAR

                  Slide 76 in the PDF

                  SHAR (Patton & Sheppard, 2015) splits the daily term by the sign of the intraday returns, \(\mathrm{RV}_{t+1:t+h} = \beta_0 + \beta^{+}\mathrm{RV}_t^{+} + \beta^{-}\mathrm{RV}_t^{-} + \beta_w\mathrm{RV}_t^{(w)} + \beta_m\mathrm{RV}_t^{(m)} + \varepsilon_{t+h}\), where \(\mathrm{RV}_t^{+}\) and \(\mathrm{RV}_t^{-}\) are the positive and negative semivariances; \(\beta^{+} = \beta^{-}\) gives back HAR. Patton and Sheppard find \(\beta^{-} > \beta^{+}\), an intraday version of the leverage effect (Chapter 5), and the signature can capture a related effect from daily data through the Lévy area between returns and log RV.

                  In brief

                  • SHAR Patton & Sheppard (2015): split the daily term by the sign of the intraday returns
                    \begin{equation*} \RV_{t+1:t+h} = \beta_0 + \beta^{+}\RV_t^{+} + \beta^{-}\RV_t^{-} + \beta_w\RV_t^{(w)} + \beta_m\RV_t^{(m)} + \varepsilon_{t+h} \end{equation*}
                    • \(\RV_t^{+}\), \(\RV_t^{-}\): the positive and negative semivariances of day \(t\)
                    • \(\beta^{+} = \beta^{-}\) gives back HAR, since \(\RV_t^{+} + \RV_t^{-} = \RV_t\)
                  • Patton and Sheppard find \(\beta^{-} > \beta^{+}\)
                    • Variance from down-moves predicts more future variance than variance from up-moves: an intraday version of the leverage effect (Chapter 5)
                  • The signature can capture a related effect from daily data
                    • the Lévy area between returns and log RV (Path Signatures section)

                  Two Windows of JPM as Paths (1/2)

                  Slide 77 in the PDF

                  The chart draws two 22-day windows of JPM as paths in the plane of cumulative open-to-close return and log RV, with the chord from start to end and the shaded Lévy area. In the calm window to 15 November 2019 log RV moves by −0.58 and the Lévy area is −0.011; in the stress window to 16 March 2020 log RV rises by +4.32, RV multiplied by 75, while the intraday return path falls and recovers, and the Lévy area is −0.247.

                  In brief

                  Two Windows of JPM as Paths (1/2)
                  • Path (cumulative open-to-close return, log RV) over 22 days
                    • dashed chord from start to end
                    • the shaded region gives the Lévy area
                  • Calm window (to 15 Nov 2019) vs stress window (to 16 Mar 2020)
                    • calm: log RV moves by -0.58, the path wanders, Lévy area -0.011
                    • stress: log RV rises by +4.32 (RV \(\times\) 75) while the intraday return path falls and then recovers
                    • stress: Lévy area -0.247 (returns in decimal units for the area, percent on the axis)

                  Two Windows of JPM as Paths (2/2)

                  Slide 78 in the PDF

                  The net intraday return of the stress window is only +0.5%, because the crash happened largely overnight, outside the open-to-close channel. No signature level recovers what the input omits.

                  In brief

                  • The net intraday return is only +0.5%
                    • the crash was largely overnight, outside this channel
                    • no signature level recovers what the input omits
                  MFM_ch20_signatures

                  Why Signatures Work (1/2): Uniqueness and Shuffles

                  Slide 79 in the PDF

                  By the uniqueness theorem of Hambly & Lyons (2010), the full signature determines the path up to tree-like equivalence and translation, and with a strictly increasing time channel only translation remains; a finite truncation is not injective. The shuffle identity, for example \(S^iS^j = S^{ij} + S^{ji}\), writes products of signature terms as sums of higher terms, so a linear model in the truncated signature already contains polynomials of the increments.

                  In brief

                  • Uniqueness Hambly & Lyons (2010): the full signature identifies the path
                    • \(\Sig(X)\) determines \(X\) up to tree-like equivalence (adding pieces that are retraced backwards) and translation
                    • With a strictly increasing time channel nothing can be retraced, so only translation remains
                    • A finite truncation \(\Sig^N\) is not injective: two different paths can share levels \(1..N\)
                  • Shuffle identity: a product of signature terms is a sum of higher terms
                    \begin{equation*} S^{i}S^{j} = S^{ij} + S^{ji}, \qquad S^{ij}S^{k} = S^{ijk} + S^{ikj} + S^{kij} \end{equation*}
                    • The sum runs over all interleavings (shuffles) of the two words that keep the order inside each word
                    • So a linear model in \(\Sig^N(X)\) already contains polynomials of the increments

                  Why Signatures Work (2/2): Universality and Decay

                  Slide 80 in the PDF

                  Universality states that any continuous function of the path on a compact set \(K\) of time-augmented paths is approximated uniformly by a linear functional \(\langle \ell, \mathrm{Sig}(X)\rangle\) of the signature, a consequence of the Stone–Weierstrass theorem through the shuffle identity. This allows nonlinear approximation at sufficient depth but guarantees nothing at depth 3 or out of sample. Factorial decay, \(\|S^{(k)}\| \le \|X\|_{1\text{-var}}^k/k!\), makes the high levels small for paths of moderate length.

                  In brief

                  • Universality: any continuous function of the path is approximately linear in the signature
                    • Stone–Weierstrass, via the shuffle identity
                    \begin{equation*} \forall \varepsilon > 0\;\; \exists\, \ell \text{ linear}: \quad \sup_{X \in K} \big|f(X) - \langle \ell, \Sig(X)\rangle\big| < \varepsilon \end{equation*}
                    • \(K\): a compact set of time-augmented paths with a common start (in the 1-variation topology); \(f: K \to \mathbb{R}\): a continuous function of the path
                    • \(\langle \ell, \Sig(X)\rangle\): a weighted sum of finitely many signature terms, i.e. a linear regression on them
                    • Sufficient depth permits nonlinear approximation; this does not guarantee accuracy at depth 3 or out of sample. The cost is \(d^N\) features, so sparsity is needed
                  • Factorial decay: high levels are small
                    \begin{equation*} \|S^{(k)}\| \le \frac{\|X\|_{1\text{-var}}^k}{k!} \end{equation*}
                    • \(\|X\|_{1\text{-var}}\): total length of the path (sum of the lengths of its segments); \(\|S^{(k)}\|\): size (Euclidean norm) of level \(k\)
                    • Truncating at a small depth \(N\) loses little for paths of moderate length, which is why channels are rescaled (Appendix)

                  How Many Features? The Truncation Dimension

                  Slide 81 in the PDF

                  A \(d\)-dimensional path truncated at depth \(N\) has

                  \[\dim(d, N) = \sum_{k=1}^{N} d^k = \frac{d^{N+1} - d}{d - 1}\]
                  (20.23)

                  terms. With \(d = 3\) channels (time, log RV, cumulative return) and \(N = 3\) this gives \(3 + 9 + 27 = 39\) terms, of which 3 are constant and 36 vary; \(d = 5\) and \(N = 4\) would give 780 terms for 1,000 observations, too many for a stable LASSO selection. \(N = 3\) and \(L = 22\) are fixed in advance, with \(N = 2\) as a robustness check.

                  In brief

                  How Many Features? The Truncation Dimension
                  • Number of terms in levels \(1..N\) of a \(d\)-dimensional path (a geometric sum, since level \(k\) has \(d^k\) words)
                    • \(\dim(d, N) = \sum_{k=1}^{N} d^k = \frac{d^{N+1} - d}{d - 1}\)
                  • This chapter: \(d = 3\) channels (time, log RV, cumulative return), depth \(N = 3\): \(3 + 9 + 27 = 39\) terms
                    • Three are constant (words with only the time letter), so 36 vary
                  • \(d = 5\), \(N = 4\) would give 780 terms for 1000 observations: the LASSO selection would become unstable among many correlated terms
                  • Here \(N = 3\) and \(L = 22\) are fixed ex ante, with \(N = 2\) as a robustness check
                  MFM_ch20_signatures

                  Two Invariances and the Basepoint

                  Slide 82 in the PDF

                  The signature is translation invariant, \(\mathrm{Sig}(X + c) = \mathrm{Sig}(X)\), so the level of volatility enters separately through the log-HAR terms, which play the role of the external factors of Gu et al. (2024). Basepoint augmentation (Morrill et al., 2020) prepends the origin to the path, which makes the higher-order terms depend on the absolute level and the signature level-aware; translating the path to start at 0 does not, because it is itself a translation. Without a time channel the signature is also invariant to reparametrisation, so the same move run at different speeds has the same signature.

                  In brief

                  • Translation invariance: \(\Sig(X + v) = \Sig(X)\) for any constant vector \(v \in \mathbb{R}^d\)
                    • only increments enter
                    • The signature does not see the level of volatility, so the level enters separately: the log-HAR terms play the role of the external factors \(x_\tau\) of Gu et al.
                  • Basepoint augmentation Morrill et al. (2020): the path starts at the origin, then runs through the points of the window
                    \begin{equation*} \hat X = \big(0,\; X_{u_0},\; X_{u_1}, \dots, X_{u_{L-1}}\big), \qquad \Sig(\hat X) = \exp_{\otimes}(X_{u_0}) \otimes \Sig(X) \end{equation*}
                    • Level 1 becomes \(X_{u_0} + (X_{u_{L-1}} - X_{u_0}) = X_{u_{L-1}}\): the starting point does not appear on its own, but the higher-order terms depend on the absolute level (for example \(\ln\RV\)), so the signature becomes level-aware
                    • Translating the path to start at 0 does not do this: it is itself a translation (Seminar 20, C2 uses the prepended segment)
                  • Reparametrisation invariance: \(X_{\psi(u)}\), with \(\psi\) increasing, has the same signature
                    • without a time channel, two moves with the same spatial shape, run at different speeds, have the same signature
                    • the time channel tells them apart by referring them to a fixed clock; the augmented path stays invariant to a reparametrisation of the whole augmented path
                    • \(u_i = i/(L-1)\) maps every window to \([0, 1]\): the absolute duration (22 vs 2 days) needs physical time or a separate regressor
                  • In this chapter: time augmentation, no basepoint
                    • so the kernel compares shapes, not levels
                    • the consequence shows in the kernel weights (Results section)

                  Three Windows: Weights and Information

                  Slide 83 in the PDF

                  With three windows at distances \(\delta = (0.1, 0.4, 0.9)\) and \(\gamma = 2\), the unnormalised weights are \((0.819, 0.449, 0.165)\), with sum 1.433, and the normalised weights \(w = (0.571, 0.313, 0.115)\). The effective sample size \(n_{\mathrm{eff}} = 1/\sum w_i^2 \approx 2.28\), against 3 under equal weights; \(\gamma\) acts as an inverse temperature, so a larger \(\gamma\) concentrates the weights, gaining similarity and losing information.

                  In brief

                  • Distances \(\delta=(0.1, 0.4, 0.9)\) and \(\gamma=2\)
                  • Unnormalised weights: \((0.819, 0.449, 0.165)\), with sum \(1.433\)
                  • Normalised weights: \(w=(0.571, 0.313, 0.115)\)
                  • \(n_{\mathrm{eff}}=1/\sum w_i^2\approx2.28\), compared with 3 under equal weights
                  • \(\gamma\) acts as an inverse temperature
                    • a larger \(\gamma\) concentrates the weights, gaining similarity and losing information

                  Evaluation Protocol: the HAC Variance

                  Slide 84 in the PDF

                  The loss differential \(d_t\) is autocorrelated, through overlapping targets up to lag \(h - 1\) and volatility clustering in the losses, so the plain variance understates the uncertainty of \(\bar d\). The Newey–West heteroskedasticity and autocorrelation consistent (HAC) estimator (Newey & West, 1987) is

                  \[\hat\Omega = \hat\gamma_0 + 2\sum_{j=1}^{q}\Big(1 - \frac{j}{q+1}\Big)\hat\gamma_j,\]
                  (20.24)

                  with \(\hat\gamma_j\) the sample autocovariance of \(d_t\) at lag \(j\); the Bartlett weights guarantee \(\hat\Omega \ge 0\), and \(q = \max(h, \lceil T^{1/3}\rceil)\) lags cover at least the overlap.

                  In brief

                  • Why not the plain sample variance: \(d_t\) is autocorrelated
                    • overlapping targets up to lag \(h - 1\)
                    • volatility clustering in the losses
                    • so \(\hat\Gamma_0/T\) understates the uncertainty of \(\bar d\) (and over-rejects) when the sum of the autocovariances is positive, the typical case with overlapping targets
                  • The Newey–West estimator Newey & West (1987), heteroskedasticity and autocorrelation consistent (HAC)
                    \begin{equation*} \hat\Omega = \hat\Gamma_0 + 2\sum_{j=1}^{q}\Big(1 - \frac{j}{q+1}\Big)\hat\Gamma_j, \qquad \hat\Gamma_j = \frac{1}{T}\sum_{t=j+1}^{T}(d_t - \bar d)(d_{t-j} - \bar d) \end{equation*}
                    • \(\hat\Gamma_j\): the sample autocovariance of \(d_t\) at lag \(j\) (not the kernel temperature \(\gamma\))
                    • The Bartlett weights \(1 - j/(q+1)\) decline linearly with the lag and guarantee \(\hat\Omega \ge 0\)
                    • \(q = \max(h, \lceil T^{1/3}\rceil)\) lags: at least \(h\), for the dependence created by the overlap (\(\lceil\cdot\rceil\): rounding up)

                  Two Numerical Examples: Holm, BH and the MCS

                  Slide 85 in the PDF

                  Holm compares three sorted p-values, \((0.004, 0.023, 0.08)\), in turn with \((0.05/3, 0.05/2, 0.05)\): the first two pass, the third fails, and the procedure stops. BH compares them with \((0.05/3, 2\cdot0.05/3, 0.05)\) and keeps the largest passing rank, again two; in a family of 50 tests the same p-values may no longer pass. For an MCS with models A, B and C and standardised loss gaps 2.4, −0.5 and −0.7, the bootstrap p-value 0.04 removes A at 10%, the retest of B and C gives 0.30, and both remain, which does not prove that their population losses are equal.

                  In brief

                  • Holm compares three sorted p-values, \((0.004, 0.023, 0.08)\), in turn with \((0.05/3, 0.05/2, 0.05)\)
                    • the first two pass, the third fails, so we stop
                  • BH compares them with \((0.05/3, 2 \cdot 0.05/3, 0.05)\) and keeps the largest passing rank: again two
                  • In a family of 50 tests the same p-values may no longer pass
                    • which is why the family is fixed before the results
                  • MCS with three models A, B and C, standardised loss gaps 2.4, \(-0.5\) and \(-0.7\)
                    • the bootstrap p-value is 0.04, so at 10% we remove A
                  • Recompute the test for B and C only: the new p-value is 0.30, so we stop and B and C remain
                  • Failure to separate them is not proof that their population losses are equal
                  • The 90% coverage is a repeated-sample guarantee, not a posterior probability for the set obtained

                  DM Statistics for Every Asset and Horizon

                  Slide 86 in the PDF

                  The heat map gives the DM \(t\)-statistics of the QLIKE differential of Sig-LK for each of the 40 stocks, 5 futures and 5 FX pairs, blue where Sig-LK is more accurate and red where it is less. Against log-HAR the map is pale at \(h = 1\) and red at \(h = 5\) and 22, with blue exceptions for ES, Tesla (TSLA) and EURUSD and five assets significantly red before correction; against HAR it is blue at \(h = 1\), because Sig-LK inherits the log target.

                  In brief

                  DM Statistics for Every Asset and Horizon
                  • Diebold–Mariano \(t\)-statistics of the QLIKE differential, one column per asset ticker
                    • 40 stocks, 5 futures, 5 FX pairs
                    • blue: Sig-LK more accurate, red: less accurate
                  • Against log-HAR (top): pale at \(h = 1\), red at \(h = 5, 22\)
                    • blue exceptions: ES, TSLA = Tesla, EURUSD
                    • five assets significantly red before correction
                  • Against HAR (bottom): blue at \(h = 1\) because Sig-LK inherits the log target
                  MFM_ch20_case_study

                  Does the Kernel Ever Favour Stress? (1/2)

                  Slide 87 in the PDF

                  The chart shows the weight that Sig-LK (\(h = 1\)) puts on the 10% most volatile days of its training window, averaged by class over three months. For stocks the mean is 6.7% per re-estimation, below the 10% of equal weighting, ranging from 4.4% to 9.9% (maximum on 17 March 2020); FX stays below 10% throughout, and futures exceed it only in early 2016 (up to 11.7%) and in March–April 2020.

                  In brief

                  Does the Kernel Ever Favour Stress? (1/2)
                  • Weight that Sig-LK (\(h = 1\)) puts on the 10% most volatile days of its training window
                    • averaged by class, three-month mean
                  • Stocks, per re-estimation: mean 6.7%, below the 10% of equal weighting
                    • between 4.4% and 9.9% (maximum on 17 Mar 2020)
                  • FX stays below 10% throughout
                  • Futures exceed 10% (three-month mean) only in two short episodes
                    • early 2016 (up to 11.7%)
                    • March–April 2020

                  Does the Kernel Ever Favour Stress? (2/2)

                  Slide 88 in the PDF

                  This is the mechanism behind the null result for H2: the weights do not implement "learn from past crises".

                  In brief

                  • The mechanism behind the null result for H2
                    • the weights do not implement “learn from past crises”
                  MFM_ch20_case_study

                  What the LASSO Selects

                  Slide 89 in the PDF

                  Across all assets and horizons, the LASSO, with its support selected by BIC, keeps the three log-HAR terms in 99%, 94% and 93% of the re-estimations, so it rediscovers HAR. The average support has 8.7 regressors for Sig-L and 6.7 for Sig-LK. The most frequent signature term, (t, r, r), selected in 27% of the re-estimations, and all the frequent ones are of level 3, cubic in the increments.

                  In brief

                  What the LASSO Selects
                  • Across all assets and horizons: share of re-estimations in which a regressor is in the support of Sig-LK
                    • the support is selected by BIC
                    • labels: HAR day/week/month = the three log-HAR terms, t = time, ln RV = \(\ln\RV\), r = cumulative return
                  • The log-HAR terms are kept in 99%, 94% and 93% of the re-estimations
                    • the LASSO rediscovers HAR
                    • average support 8.7 regressors (Sig-L), 6.7 (Sig-LK)
                  • The most frequent signature term, (t,r,r) (27%), and all frequent ones are of level 3
                    • cubic in the increments
                    • far outside the training range in a crash, amplified by \(\exp(\cdot)\)
                  MFM_ch20_case_study

                  Robustness of Sig-LK

                  Slide 90 in the PDF

                  The chart compares Sig-LK relative to HAR, as a mean over the 50 assets, in the main design, in two pre-registered variants and with the BPQ filter; the reference line marks log-HAR, and a bar below 1 means a more accurate forecast than HAR.

                  In brief

                  Robustness of Sig-LK
                  • Sig-LK relative to HAR (mean over 50 assets)
                    • in the main design and in two pre-registered variants
                    • with the BPQ filter (the disclosed deviation)
                  • The reference line marks log-HAR
                    • a bar below 1 means a more accurate forecast than HAR
                  MFM_ch20_case_study

                  Robustness of Sig-LK: the Variants

                  Slide 91 in the PDF

                  At depth 2 the ratios at \(h = 5\) and 22 fall from 1.104 and 1.334 in the main design to 0.922 and 0.820, and without the return channel to 0.878 and 0.770, close to log-HAR (0.871 and 0.765). The ablations suggest that the long-horizon failures are linked to level-3 terms that contain the return channel, although removing terms also changes the selection, the kernel distances and the weights. Both variants reduce the instability; neither establishes H1.

                  In brief

                  • Depth 2: 0.922 and 0.820 at \(h = 5, 22\) (main design: 1.104, 1.334)
                  • Path without returns: 0.878 and 0.770
                    • the ablations suggest that the long-horizon failures are linked to terms that contain the return channel
                  • Depth 2 also reduces them: the ablations suggest a role of the level-3 terms
                    • removing terms, however, also changes the selection, the kernel distances and the weights
                  • Without returns the model approaches log-HAR (0.871 and 0.765)
                    • depth 2 trails it
                  • Both variants reduce instability
                    • neither establishes H1

                  Appendix

                  The Appendix gives the details of the features, the estimation, the evaluation windows and the chosen temperatures, the paper's own error metric on VOLARE, the filtered results, the researcher's checklist and the prompts of the discovery loop.

                  Features and Scaling

                  Slide 92 in the PDF

                  The regressors at day \(\tau\) are the three log-HAR terms and the 39 terms of the depth-3 signature of the 22-point path (time, \(\ln\mathrm{RV}\), cumulative return). Each non-time channel \(q\) is multiplied by \(\kappa_q = 1/(s_q\sqrt{L-1})\), with \(s_q\) the standard deviation of its daily increments in the training window, so its typical total move over the window is about 1; time runs from 0 to 1, a word is rescaled by the product of its factors, every column is standardised with weighted moments and constant columns are dropped.

                  In brief

                  Route Decay Kernel
                  • Regressors at day \(\tau\) for the signature models: 3 log-HAR terms + 39 signature terms
                    • \(\xi_\tau = \big[\ln\RV_\tau,\; \ln\RV^{(w)}_\tau,\; \ln\RV^{(m)}_\tau,\; \Sig^3(X_{\tau-21:\tau})\big]\), path \(X\) = (time, \(\ln\RV\), cumulative \(r\)), 22 points
                  • Each non-time channel \(q\) is multiplied by \(\kappa_q = 1/(s_q\sqrt{L - 1})\)
                    • so its typical total move over the window is about 1
                    • \(s_q\): standard deviation of the daily increments of channel \(q\) from the start of the training window up to the origin \(t\) (\(W + h\) days)
                    • \(L - 1 = 21\): the number of increments in one path, used only in the factor \(\sqrt{L - 1}\)
                    • Time is not rescaled (constant increments, \(s = 0\)): it runs from 0 to 1; a word \((i_1..i_k)\) is rescaled by \(\prod_j \kappa_{i_j}\)
                  • Before the LASSO every column is standardised with weighted moments of the training window
                    • constant columns (words of time only) are dropped
                  • Raw features at \(\tau\) use data up to \(\tau\)
                    • scales and standardisation use data up to the origin \(t\)
                    • the no-look-ahead cutoff is \(t\)

                  Estimation: Choosing \(\lambda\) Without Look-Ahead

                  Slide 93 in the PDF

                  Inside each training window the whole weighted LASSO path is computed by least-angle regression (LARS), each support on the path is refitted by weighted OLS (step 2 of Gu et al. (2024)), and the support minimising

                  \[\mathrm{BIC}(S) = n_{\mathrm{eff}}\ln \hat s^2_S + (|S| + 1)\ln n_{\mathrm{eff}}, \qquad n_{\mathrm{eff}} = 1\Big/\sum_\tau w_\tau^2,\]
                  (20.25)

                  is chosen, with \(\hat s^2_S\) the weighted residual variance. Cross-validation would be too costly over 50 assets, 3 horizons and hundreds of re-estimations; the Kish size \(n_{\mathrm{eff}}\) corrects for unequal weights but not for serial dependence, so (20.25) is a heuristic of the design, not a likelihood-based BIC.

                  In brief

                  • Inside each training window: the whole weighted LASSO path by LARS
                    • For every support on the path: weighted OLS refit (step 2 of Gu et al.)
                  • Pick the support that minimises a BIC-type criterion with the Kish effective sample size
                    • a heuristic of this design
                    • \(\mathrm{BIC}(S) = n_{\mathrm{eff}}\ln \hat s^2_S + (|S| + 1)\ln n_{\mathrm{eff}}\), \(\quad n_{\mathrm{eff}} = 1/\sum_\tau w_\tau^2\)
                    • \(S\): a support (set of kept regressors), \(|S|\): its size, \(+1\) for the intercept; \(\hat s^2_S\): weighted residual variance of the refit on \(S\)
                  • Why not cross-validation: 50 assets \(\times\) 3 horizons \(\times\) hundreds of re-estimations
                    • BIC uses only the window
                    • Kish size corrects for unequal weights, not for serial dependence (overlapping targets, \(h > 1\)): not a likelihood-based BIC for these data

                  Evaluation Windows

                  Slide 94 in the PDF

                  Evaluation starts on 27 January 2020 for stocks, 25 September 2014 for futures and 27 August 2014 for FX, and ends on 29 January 2026, with median lengths of 1,499, 2,919 and 2,952 days per asset. Each model produces 265,929 forecasts over the three horizons: 89,893 at \(h = 1\), 89,293 at \(h = 5\) and 86,743 at \(h = 22\). Because the stock data start in 2015, the COVID-19 crash is in every stock evaluation sample.

                  In brief

                  ClassEvaluation startsEndsDays per asset (median)
                  Stocks27 Jan 202029 Jan 20261 499
                  Futures25 Sep 201429 Jan 20262 919
                  FX27 Aug 201429 Jan 20262 952
                  • 265 929 forecasts per model over the three horizons: 89 893 (\(h = 1\)), 89 293 (\(h = 5\)), 86 743 (\(h = 22\))
                  • Stocks start in 2015, so their evaluation begins in January 2020
                    • the COVID-19 crash is in every stock sample
                    • Futures and FX start in 2009 and are evaluated from 2014

                  Chosen Kernel Temperatures

                  Slide 95 in the PDF

                  The kernel temperature \(c\) is chosen from the grid {0.5, 1, 2, 4} on the validation block of each class and horizon. For Sig-LK it is 0.5 for stocks at all horizons, 2.0, 1.0 and 0.5 for futures, and 0.5, 2.0 and 0.5 for FX; log-HAR-K uses the largest value, 4.0, for futures at \(h = 22\) and for FX at \(h = 1\) and 22.

                  In brief

                  Sig-LKlog-HAR-K
                  Class\(h = 1\)\(h = 5\)\(h = 22\)\(h = 1\)\(h = 5\)\(h = 22\)
                  Stocks0.50.50.50.50.51.0
                  Futures2.01.00.52.01.04.0
                  FX0.52.00.54.02.04.0
                  • The value of \(c\) from the grid \(\{0.5, 1, 2, 4\}\), chosen on the validation block of each class and horizon

                  The Paper’s Metric on VOLARE

                  Slide 96 in the PDF

                  On the paper's own metric, the median over assets of the mean relative error \(|\mathrm{RV} - F|/\mathrm{RV}\), log-HAR has 50.4%, 41.9% and 38.7% at \(h = 1\), 5 and 22, against 57.2%, 48.9% and 51.7% for HAR, 50.1%, 42.6% and 42.6% for Sig-L and 50.6%, 43.6% and 43.7% for Sig-LK.

                  In brief

                  Median of asset mean errors \(|\RV - F|/\RV\) (%)\(h = 1\)\(h = 5\)\(h = 22\)
                  HAR57.248.951.7
                  log-HAR50.441.938.7
                  HARQ55.049.050.7
                  SHAR57.949.251.8
                  log-HAR-K49.741.038.7
                  Sig-L50.142.642.6
                  Sig-LK50.643.643.7
                  • Median over assets of the mean relative error: about 40–60%
                    • against 1–2% for freight rates in Gu et al.
                  • Daily RV is a noisy, heavy-tailed target
                    • the evaluated models have much larger relative errors than those reported for freight rates
                    • this does not establish a lower bound on attainable error
                    • Sig-LK forecasts above ten times the in-sample maximum: 8 at \(h = 1\), 29 at \(h = 5\) (concentrated on 10 assets) and 36 at \(h = 22\); log-HAR: none

                  The Insanity Filter, Revisited

                  Slide 97 in the PDF

                  With the BPQ filter applied to all models, the QLIKE ratios to HAR at \(h = 1\), 5 and 22 are 0.763, 0.654 and 0.778 for log-HAR, 0.995, 0.912 and 1.014 for HARQ, 0.871, 0.819 and 0.904 for Sig-L and 0.890, 0.867 and 0.930 for Sig-LK. The filter rescues HARQ, but log-HAR keeps the lowest loss at every horizon.

                  In brief

                  QLIKE relative to HAR, BPQ filter on all models\(h = 1\)\(h = 5\)\(h = 22\)
                  HAR1.0001.0001.000
                  log-HAR0.7630.6540.778
                  HARQ0.9950.9121.014
                  SHAR1.0141.0031.000
                  log-HAR-K0.7720.6590.779
                  Sig-L0.8710.8190.904
                  Sig-LK0.8900.8670.930
                  • The filter rescues HARQ relative to HAR (0.995 at \(h = 1\) instead of 1.140), as BPQ intended
                  • Sig-LK relative to log-HAR, both filtered: 1.238, 1.424, 1.209
                    • the gap widens at \(h = 1, 5\) and narrows at \(h = 22\)
                  • Filtered forecasts, \(h = 1\): HAR 180, HARQ 140, Sig-LK 143, log-HAR 29 (of 89 893)
                    • the filter hits the signature models far more often than log-HAR

                  What the Researcher Must Check

                  Slide 98 in the PDF

                  The checklist has four items:

                  1. The method must be the paper's, with each component pointing to the equation, table or section it comes from and undisclosed choices labelled as choices of the study.
                  2. There must be no look-ahead: changing every value after the origin, for example multiplying future RV by 10, must leave the forecast unchanged.
                  3. The benchmark must isolate the idea, through the 2 × 2 ablation.
                  4. The losses and tests must be the right way round, with \(\mathrm{QL}(\mathrm{RV}, \mathrm{RV}) = 0\) and a HAC variance with bandwidth at least \(h\).

                  In brief

                  • The method is the paper’s: kernel, weights, refit, as in the cited equations
                    • Test: each component of the method points to the equation, table or section of the paper it comes from; undisclosed choices are labelled as choices of the study
                  • No look-ahead: \(\lambda\), scales, \(c\) and features use only data available at the origin
                    • Test: change every value after the origin (for example, multiply future RV by 10); the forecast made at the origin must not change
                  • The benchmark isolates the idea: the \(2 \times 2\) ablation is reported, not only the full model against HAR
                  • Losses and tests are the right way round: \(\QL(\RV, \RV) = 0\) and \(\QL(\RV, F) > 0\) otherwise
                    • The DM variance is HAC with bandwidth at least \(h\)

                  The Discovery Loop, Prompts (1/2)

                  Slide 99 in the PDF

                  The reading prompt asks the assistant to list every equation of the estimator of Gu et al. (2024), every hyperparameter with its reported value and the benchmark of each table, marking what is not disclosed, and each item is checked against the PDF. The mapping prompt asks, for daily realised variance, whether each component transfers, what must change and which benchmark isolates it; the pre-registration is then fixed in a dated document signed by the team before the first result.

                  In brief

                  • Read: extract exactly what the paper did
                    • “From Gu et al. (2024, KDD), list every equation of the estimator, every hyperparameter and its reported value, and the benchmark of each table. Mark what is not disclosed.”
                    • Check each item against the PDF: equation numbers, table numbers, the kernel actually used
                  • Map: what transfers, what must be adapted
                    • “Our target is daily realised variance. For each component of the method, say whether it transfers, what must change, and which benchmark isolates it.”
                  • Pre-register: hypotheses, grid, tests, correction
                    • fixed in a dated document signed by the team before the first result

                  The Discovery Loop, Prompts (2/2)

                  Slide 100 in the PDF

                  The testing prompt asks for unit tests of a depth-3 signature: level 1 equal to the increment, the symmetric part of level 2 equal to half the outer product, and Chen's identity on a split path. The attack prompt gives all 600 runs, 50 assets by 3 horizons by 4 specifications, and asks which conclusions survive Holm over 50 assets and which single asset or subperiod would make a misleading headline; the last step explains the null by its mechanism, not by an excuse.

                  In brief

                  • Build and test the pieces: signatures against identities, not against another library
                    • “Write unit tests for a depth-3 signature: level 1 equals the increment, the symmetric part of level 2 equals half the outer product, Chen’s identity on a split path.”
                  • Run the grid, then attack it
                    • “Here are all 600 runs (50 assets \(\times\) 3 horizons \(\times\) 4 specifications: main, depth 2, no return channel, BPQ filter). Which conclusions survive Holm over 50 assets? Which single asset or subperiod would make a misleading headline?”
                  • Explain the null: find the mechanism, not an excuse

                  Acronyms used in this material (1/2)

                  Slide 101 in the PDF
                  • AAPL: Apple Inc. (Nasdaq ticker)
                  • ACM: Association for Computing Machinery
                  • AI: Artificial Intelligence
                  • AR: AutoRegressive model
                  • ARIMA: AutoRegressive Integrated Moving Average (model)
                  • AUD: Australian Dollar (ISO 4217 code)
                  • BH: Benjamini–Hochberg (procedure)
                  • BIC: Bayesian Information Criterion
                  • BPQ: Bollerslev–Patton–Quaedvlieg (insanity filter of the HARQ paper)
                  • CAD: Canadian Dollar (ISO 4217 code)
                  • CL: Crude oil (WTI light sweet) futures, NYMEX (New York Mercantile Exchange) ticker
                  • COVID-19: COronaVIrus Disease 2019
                  • DM: Diebold–Mariano (test)
                  • ES: E-mini S&P 500 futures (CME ticker)
                  • ETF: Exchange-Traded Fund
                  • EURUSD: Euro / US dollar exchange rate (FX pair code)
                  • FDR: False Discovery Rate
                  • FWER: Family-Wise Error Rate
                  • FX: Foreign eXchange
                  • GARCH: Generalised AutoRegressive Conditional Heteroskedasticity
                  • GC: Gold futures, COMEX (Commodity Exchange, metals futures, CME Group) ticker
                  • GW: Giacomini–White (test)
                  • HAC: Heteroskedasticity and Autocorrelation Consistent (standard errors)

                  Acronyms used in this material (2/2)

                  Slide 102 in the PDF
                  • HAR: Heterogeneous AutoRegressive (model)
                  • HARQ: HAR model with Realised Quarticity
                  • JPM: JPMorgan Chase & Co. (NYSE ticker)
                  • KDD: Knowledge Discovery and Data Mining (ACM SIGKDD conference)
                  • LARS: Least-Angle RegreSsion (algorithm)
                  • LASSO: Least Absolute Shrinkage and Selection Operator
                  • MCS: Model Confidence Set
                  • MSE: Mean Squared Error
                  • NG: Natural Gas futures, NYMEX ticker
                  • NVDA: NVIDIA Corporation (Nasdaq ticker)
                  • NW: Newey–West (standard errors)
                  • NYSE: New York Stock Exchange
                  • OLS: Ordinary Least Squares
                  • PDE: Partial Differential Equation
                  • PRDS: Positive Regression Dependence on a Subset (of p-values)
                  • QLIKE: Quasi-LIKElihood (loss function)
                  • RQ: Realised Quarticity
                  • RV: Realised Variance
                  • SHAR: Semivariance HAR model
                  • TSLA: Tesla, Inc. (Nasdaq ticker)
                  • US: United States
                  • USDJPY: US dollar / Japanese yen exchange rate (FX pair code)
                  • VOLARE: VOLatility Archive for Realized Estimates (database)
                  • WTI: West Texas Intermediate (crude oil benchmark)
                  • XOM: Exxon Mobil Corporation (NYSE ticker)

                  Realised variance from the VOLARE database (Cipollini et al., 2026): 40 US stocks (2015–2026), 5 futures and 5 currency pairs (2009–2026), five-minute returns. VOLARE data are licensed for research use and are not redistributed; only summaries appear here: per-asset mean losses and MCS p-values, kernel weights and distances, five-day means of the volatility of the six assets on the slide (one value per week), and the JPM forecasts of February–May 2020. The lead-lag case study uses the five-minute bars of SPY and Bitcoin (EODHD, data/market/intraday): the cross-correlations and mean Lévy areas of each day.

                  References