Paper: arXiv 2506.11843

Authors: Emmanouil Sfendourakis

Abstract

We present a Markovian market model driven by a hidden Brownian efficient price. In particular, we extend the queue-reactive model, making its dynamics dependent on the efficient price. Our study focuses on two sub-models: a signal-driven price model where the mid-price jump rates depend on the efficient price and an observable signal, and the usual queue-reactive model dependent on the efficient price via the intensities of the order arrivals. This way, we are able to correlate the evolution of limit order books of different stocks. We prove the stability of the observed mid-price around the efficient price under natural assumptions. Precisely, we show that at the macroscopic scale, prices behave as diffusions. We also develop a maximum likelihood estimation procedure for the model, and test it numerically. Our model is them used to backest trading strategies in a liquidation context.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a complex Markovian framework with advanced stochastic calculus, proofs of stability, and a PDE-based likelihood approximation, indicating high mathematical density. Empirically, it includes a numerical testing procedure, a real-data application, and a backtest for liquidation strategies, giving it substantial practical validation.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Unify Queue-Reactive &<br>Signal-Driven LOB Models"] --> B["Model Formulation<br>Hidden Brownian Efficient Price<br>Multi-dimensional Dynamics"]
  B --> C["Theoretical Analysis<br>Prove Mid-Price Stability &<br>Macroscopic Diffusive Behavior"]
  B --> D["Estimation & Testing<br>Maximum Likelihood Estimation<br>Numerical Validation"]
  C & D --> E["Application<br>Backtesting Trading Strategies<br>in Liquidation Context"]
  E --> F["Key Outcomes<br>Unified Framework<br>Cross-Asset Correlation<br>Validated Methodology"]

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 Markovian market model driven by a hidden Brownian efficient price that extends the queue‑reactive model and includes a signal‑driven price sub‑model.

Source sentence
We present a Markovian market model driven by a hidden Brownian efficient price.

Assumes 6

  • The efficient price follows a Brownian motion and the observed mid‑price is stable around it under natural assumptions. model
    Source sentence
    We prove the stability of the observed mid-price around the efficient price under natural assumptions.
  • Intensities of order arrivals are bounded when queue sizes exceed a threshold (Assumption 3.2.2). model
    Source sentence
    9 3.2.2 Results We start stating some assumptions that will ensure that finite volumes stay in the LOB, and that its dynamics follow the efficient price.
  • Arrival intensity of large orders remains bounded (Assumption 3.2.3). model
    Source sentence
    There exist Q ∈N∗and λ ∈(0, ∞) such that for i ∈I, q = (qj)j∈J ∈Q, j ∈J and y ∈R, if qj > Q, then X e=(+,j′,s,n)∈T+ j′=j n˜Λi,e(q, y) + X e=(modif,j′,j′′,n)∈Tmodif j′′=j n˜Λi,e(q, y) ⩽λ. Assumption 3.2.3.
  • Intensities of limit order arrivals, modifications and reference price changes stay bounded as the efficient price remains at a bounded distance from the reference price (Assumption 3.2.4). model
    Source sentence
    There exists Q′ ∈N∗and λ ∈(0, ∞] such that for all i ∈I, j ∈J, q ∈Q and y ∈R, X e=(+,j′,n)∈T+ j′=j n>Q′ n˜Λi,e(q, y) + X e=(modif,j′,j′′,n)∈Tmodif j′=j n>Q′ n˜Λi,e(q, y) ⩽λ. Assumption 3.2.4.
Show 2 more
  • The connection between the efficient price and order flow is ensured by boundedness of intensity functions (Assumption 3.2.5). model
    Source sentence
    For i ∈I and e ∈T, supq∈Q ˜Λi,e(q, ·)1{δpe(q)̸=0} is locally bounded. Assumption 3.2.5.
  • The reference price jumps by one tick when the best ask or bid queue is depleted. market
    Source sentence
    When the ask (resp. bid) queue is depleted the reference price jumps up (resp. down) by one tick δ = 0.01.

Data

  • Simulated data for Model 1 (queue‑reactive, d=1) synthetic unknown
    Source sentence
    We test the two following models. Model 1: d = 1 stock.
  • Simulated data for Model 2 (signal‑driven, d=2) synthetic unknown
    Source sentence
    Model 2: d = 2 stocks. Signal-driven price model of Section 3.1, with a state space X := R2 ×R2, the first two components being associated to Stock 1, the next ones to Stock 2.

Costs

Modeled Costs are part of the reported headline results.

Transient market impact is modeled; trading cost increases with trading speed and consists of a transient part that vanishes over time and a permanent part.

Source sentence
Trading too fast is suboptimal because of transient market impact: the cost of trading is an increasing function of trading speed. In general, the market impact of a trade is defined as the difference between the actual price of the asset and the price the asset would have if the said trade never occurred (the definition of the “price” can vary: mid-price, last traded price or reference price to cite a few).

Finds 5

  • The observed mid‑price is stable around the efficient price and converges to a diffusion at the macroscopic scale. theoretical
    Source sentence
    We prove the stability of the observed mid-price around the efficient price under natural assumptions. Precisely, we show that at the macroscopic scale, prices behave as diffusions.
  • Under the stated assumptions, the scaled reference price process converges in distribution to a d‑dimensional Brownian motion. theoretical
    Source sentence
    Corollary 3.2.2. Under Assumptions 3.2.1, 3.2.2, 3.2.3, 3.2.4 and 3.2.5, ( ˜P (n))n∈N∗converges in distribution in D([0, ∞), Rd) to ΣB where B is a d-dimensional Brownian motion.
  • Estimated imbalance coefficients are consistently negative, matching empirical findings that higher ask volume predicts price declines. statistical
    Source sentence
    The imbalance coefficients αBNP,− imbalance and αSOGN,− imbalance are consistently negative, which corresponds to the know empirical findings: when there is more volume on the best ask pile than on the bid, the price is likely to go down in the near future.
  • The estimated correlation coefficient between the two stocks is always positive and high, indicating positively correlated price moves at the macroscopic scale. statistical
    Source sentence
    The correlation coefficient ρ is always reported positive (and high), validating that the price move of the two stocks are, in a macroscopic scale, positively correlated.
  • The model exhibits only transient market impact; the reference price returns close to the efficient price after a trade. performance
    Source sentence
    On the contrary, the market impact in the models presented here can only be transient: we saw that the reference price will always come back close to the efficient price.

Checked by others

Supported by

  • Liquidity Dynamics in RFQ Markets and Impact on Pricing (2023)
    On the question: Does order‑book imbalance predict subsequent price movements?
    Compare the findings

    This paper: Estimated imbalance coefficients are consistently negative, matching empirical findings that higher ask volume predicts price declines.

    That paper: Flow imbalances contain information about price evolution and can be used to generalize the micro‑price notion.

    Both findings report a negative imbalance coefficient, confirming that higher ask volume signals price declines; no market or sample differences noted.

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.

Cited References (44)

  1. Simulating and analyzing order book data: The queue-reactive model (2015)
  2. Ergodicity and diffusivity of Markovian order book models: a general framework (2017)
  3. Estimating the efficient price from the order flow: a Brownian Cox process approach (2013)
  4. A Bayesian viewpoint on the price formation process (2023) — link
  5. Stochastic stability and control (1967)
  6. Convergence of probability measures (1999)
  7. Stability of Markovian processes I: Criteria for discrete-time chains (1992)
  8. Stability of Markovian processes II: Continuous-time processes and sampled chains (1993)
and 36 more
  1. Stability of Markovian processes III: Foster–Lyapunov criteria for continuous-time processes (1993)
  2. Affine Volterra processes (2019) — link
  3. Optimal Novikov-type criteria for local martingales with jumps (2012)
  4. Partial differential equations of parabolic type (1983)
  5. Understanding the least well-kept secret of high-frequency trading — → 2307.15599
  6. Price dynamics in a Markovian limit order market (2013)
  7. Order book dynamics in liquid markets: limit theorems and diffusion approximations (2012) — link
  8. Limit order strategic placement with adverse selection risk and the role of latency (2017)
  9. Optimal liquidity-based trading tactics (2021)
  10. Modelling intensities of order flows in a limit order book (2017)
  11. LOB modeling using Hawkes processes with a state-dependent factor (2023)
  12. The CMA evolution strategy: A tutorial (2016) — link
  13. Cmaes: A simple yet practical Python library for CMA-ES (2024) — link
  14. On parameter estimation of hidden ergodic Ornstein-Uhlenbeck process (2019)
  15. Consistency of MLE in partially observed diffusion models on a torus (2025) — link
  16. Consistency of MLE for partially observed diffusions, with application in market microstructure modeling (2024) — link
  17. Limit order books (2013)
  18. M i FID II : Impact of the new tick size regime. (2018)
  19. A stochastic model for order book dynamics (2010)
  20. The predictive power of zero intelligence in financial markets (2005)
  21. Statistical theory of the continuous double auction (2003)
  22. A mathematical approach to order book modeling (2013)
  23. Long-Time Behavior of a Hawkes Process–Based Limit Order Book (2015)
  24. High frequency asymptotics for the limit order book (2016)
  25. Modelling microstructure noise with mutually exciting point processes (2013)
  26. State-dependent Hawkes processes and their application to limit order book modelling (2022)
  27. A single queue reactive Hawkes model for the order flow (2022) — link
  28. A Novel Approach to Queue-Reactive Models: The Importance of Order Sizes — → 2405.18594
  29. Deep Learning Meets Queue-Reactive: A Framework for Realistic Limit Order Book Simulation — → 2501.08822
  30. Some limit theorems for Hawkes processes and application to financial statistics (2013)
  31. Limit theorems for nearly unstable Hawkes processes (2015)
  32. Risk minimizing hedging for a partially observed high frequency data model (2006)
  33. Risk-minimizing hedging strategies under restricted information: The case of stochastic volatility models observable only at discrete random times (1999)
  34. A new approach for the dynamics of ultra-high-frequency data: the model with uncertainty zones (2011) — link
  35. From asymptotic properties of general point processes to the ranking of financial agents (2019) — link
  36. Optimal execution of portfolio transactions (2001)

3 of 44 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.