Paper: arXiv 2405.11444

Authors: Jonathan Chávez-Casillas, José E. Figueroa-López, Chuyi Yu, Yi Zhang

Abstract

A novel high-frequency market-making approach in discrete time is proposed that admits closed-form solutions. By taking advantage of demand functions that are linear in the quoted bid and ask spreads with random coefficients, we model the variability of the partial filling of limit orders posted in a limit order book (LOB). As a result, we uncover new patterns as to how the demand’s randomness affects the optimal placement strategy. We also allow the price process to follow general dynamics without any Brownian or martingale assumption as is commonly adopted in the literature. The most important feature of our optimal placement strategy is that it can react or adapt to the behavior of market orders online. Using LOB data, we train our model and reproduce the anticipated final profit and loss of the optimal strategy on a given testing date using the actual flow of orders in the LOB. Our adaptive optimal strategies outperform the non-adaptive strategy and those that quote limit orders at a fixed distance from the midprice.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper introduces complex stochastic demand functions and general price dynamics without martingale assumptions, requiring advanced mathematical modeling; it is heavily data-driven, using LOB data for model training, calibration, backtesting, and out-of-sample validation across a year-long dataset.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Design adaptive market-making strategies<br>with closed-form solutions for LOB"] --> B{"Methodology"}
  
  B --> C["Model LOB Dynamics"]
  B --> D["Estimate Demand Functions<br>Linear in spreads with random coefficients"]
  B --> E["Derive Optimal<br>Placement Strategy"]
  
  C --> F["Data Processing<br>LOB training data"]
  D --> F
  E --> F
  
  F --> G["Computation"]
  
  G --> H["Train Model<br>Parameter estimation"]
  G --> I["Simulate Strategy<br>Apply to test date"]
  
  H --> J["Key Findings<br>• Adaptive strategy outperforms fixed/ non-adaptive<br>• Closed-form solutions enable online adaptation<br>• Captures impact of demand randomness"]

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 novel high‑frequency market‑making approach in discrete time with closed‑form optimal placement strategies that adapt online to market order behavior.

Source sentence
... The most important feature of our optimal placement strategy is that it can react or adapt to the behavior of market orders online.

Assumes 5

  • Demand for limit order execution is linear in the quoted bid and ask spreads with random coefficients. model
    Source sentence
    By taking advantage of demand functions that are linear in the quoted bid and ask spreads with random coefficients, we model the variability of the partial filling of limit orders posted in a limit order book (LOB).
  • The terminal liquidation cost is modeled as IT (ST − λ IT). execution
    Source sentence
    the terminal liquidation cost or price impact, originated from the use of a market order, is modeled as IT (ST −λIT ), with ST and IT respectively denoting the final fundamental stock price and the MM’s inventory.
  • Price increments are conditionally independent of order‑flow variables and form a martingale. market
    Source sentence
    ... are conditionally independent given Ftk, and (ii) {Stk+1− Stk}k=0,...,N is a martingale, i.e., E[Stk+1 −Stk|Ftk] = 0
  • The market maker quotes at preset discrete times and liquidates inventory at the terminal time. execution
    Source sentence
    We consider an intraday high-frequency MM who quotes both bid and ask limit orders (LO) at some prespecified discrete times and liquidates her inventory at the end of the trading period.
Show 1 more
  • The linear demand model parameters (c±, p±) are constant over a day and estimated from 1‑second intervals. statistical
    Source sentence
    To give an idea of the values of µ± cmpn, we estimate those constants for each day of the 252 days of our sample

Data

Backtest sample: MSFT stock · equities · intraday

  • Microsoft (MSFT) LOB data · Nasdaq TotalView‑ITCH 5.0 order-book commercial
    2019 · Microsoft Corporation stock
    Source sentence
    illustrate our approach with real LOB data from Microsoft Corporation (stock symbol MSFT) during the year of 2019.

Recognised sources: Nasdaq TotalView-ITCH

Costs

Modeled Costs are part of the reported headline results.

A quadratic inventory liquidation penalty −λ I_T^2 is included in the PnL objective.

Source sentence
This shows the effectiveness of the liquidation penalty −λI^2_T in controlling inventory and avoiding large end of the day costs.

Finds 5

  • Adaptive optimal strategies outperform non‑adaptive and fixed‑distance strategies in out‑of‑sample tests. performance
    Source sentence
    Our adaptive optimal strategies outperform the non-adaptive strategy and those that quote limit orders at a fixed distance from the midprice.
  • The variance of the demand slope reduces the strength of inventory adjustments required by the optimal placement. theoretical
    Source sentence
    We show that the variance of the slope further reduces the strength of this adjustment. This implies that assets with more volatile demand profiles require less strict inventory adjustments.
  • Optimal placement spreads increase with the correlation between demand slope and investors’ reservation price. theoretical
    Source sentence
    ... increase with the correlation between the demand slope and investors’ reservation price.
  • When the initial inventory is zero, the expected inventory remains zero under the optimal strategy. theoretical
    Source sentence
    The following result shows that when the initial inventory is 0, then the expected inventory stays at 0 at all future times.
  • Under both martingale and non‑martingale price dynamics, the optimal strategy reduces end‑of‑day inventory more than fixed‑level benchmarks. performance
    Source sentence
    For the two optimal strategies, the end-of-day inventory is lower than the ‘Level 1’-‘Level 6’ policies.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 24 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 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.