Paper: arXiv 2307.09077

Authors: Luca Mucciante, Alessio Sancetta

Abstract

A point process for event arrivals in high frequency trading is presented. The intensity is the product of a Hawkes process and high dimensional functions of covariates derived from the order book. Conditions for stationarity of the process are stated. An algorithm is presented to estimate the model even in the presence of billions of data points, possibly mapping covariates into a high dimensional space. The large sample size can be common for high frequency data applications using multiple liquid instruments. Convergence of the algorithm is shown, consistency results under weak conditions is established, and a test statistic to assess out of sample performance of different model specifications is suggested. The methodology is applied to the study of four stocks that trade on the New York Stock Exchange (NYSE). The out of sample testing procedure suggests that capturing the nonlinearity of the order book information adds value to the self exciting nature of high frequency trading events.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a high-dimensional point process model with exponential kernels and non-linear covariate mappings (e.g., one-hot encoding), requiring advanced theoretical results on stationarity and convergence. It includes an application on NYSE stock data using a specific dataset (Lobster), a defined out-of-sample testing procedure, and discusses implementation challenges like quadratic programming, though it lacks full backtest code or live deployment details.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Estimate Hawkes process with high dimensional<br>order book covariates for large datasets"] --> B["Data Input<br>Billions of data points<br>High-frequency trading events"]
  B --> C["Methodology<br>Point process with intensity:<br>Product of Hawkes process &<br>High dimensional covariate functions"]
  C --> D["Computation<br>Scalable algorithm for large datasets<br>Convergence & Consistency proofs"]
  D --> E["Testing<br>Out-of-sample test statistic<br>Assess model specifications"]
  E --> F["Application<br>4 NYSE stocks<br>Nonlinearity adds value to self-excitation"]
  F --> G["Key Outcomes<br>Stationarity conditions established<br>Nonlinearity captures order book dynamics"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a high‑frequency event arrival model where the intensity is a Hawkes process multiplied by high‑dimensional functions of order‑book covariates, together with an algorithm for large‑scale estimation.

Source sentence
A point process for event arrivals in high frequency trading is presented. The intensity is the product of a Hawkes process and high dimensional functions of covariates derived from the order book.

Assumes 5

  • The stochastic process g0 is uniformly bounded and weakly exogenous, satisfying a stationarity condition. statistical
    Source sentence
    ... g0 is a uniformly bounded nonnegative stochastic process satisfying Condition 1 and such that |g0|∞ PL l=1 d0,l a0,l < 1.
  • The l1‑norm of the coefficient vector b is bounded (|b|1 ≤ B) to ensure consistency. model
    Source sentence
    Condition 3. The bound |b|1 ≤B is used for technical reasons to ensure that we achieve consistency under additional regularity conditions.
  • The kernel is a sum of exponential functions, guaranteeing non‑negativity and compact parameter space. model
    Source sentence
    ... the restriction on the parameter space to a compact set ensures non‑negativity of the intensity and consistent estimation.
  • One‑hot encoding discretizes covariates into dummy variables, enabling monotonicity constraints. data
    Source sentence
    One hot‑encoding essentially discretizes the variables and maps them into dummy variables. The cost of increasing the dimensionality of the problem comes at the advantage of having variables that are easy to interpret and to which linear constraints such
Show 1 more
  • Large‑scale quadratic programming can be used to estimate the model efficiently. other
    Source sentence
    We present an estimation procedure that can deal with large datasets in a relatively simple way, using quadratic programming.

Data

  • LOBSTER Level‑3 order‑book data · LOBSTER order-book commercial
    four NYSE stocks (plus S&P500 ETF as auxiliary)
    Source sentence
    All data are traded on the NYSE. The data has been accessed through the Lobster dataset (Huang and Polak, 2011). Level 3 data allows us to correctly synchronize trades and order book.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • Out‑of‑sample tests show that incorporating nonlinear order‑book information improves the predictive power of the Hawkes intensity model. predictive-accuracy
    Source sentence
    The out of sample testing procedure suggests that capturing the nonlinearity of the order book information adds value to the self exciting nature of high frequency trading events.
  • The estimation algorithm converges and its convergence rates deteriorate only logarithmically with the number of parameters. theoretical
    Source sentence
    The results showed that the convergence rates only deteriorate at a logarithmic rate with the number of parameters.
  • One‑hot encoding enables flexible modeling of covariate impacts, yielding many non‑zero coefficients across the four stocks. statistical
    Source sentence
    The total number of estimated non zero coefficients (from one‑hot encoding) for each covariate

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 22 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 7 papers in the archive (7 by independent authors)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.