Paper: arXiv 2403.12180

Authors: Boming Ning, Kiseop Lee

Abstract

Statistical arbitrage is a prevalent trading strategy which takes advantage of mean reverse property of spread of paired stocks. Studies on this strategy often rely heavily on model assumption. In this study, we introduce an innovative model-free and reinforcement learning based framework for statistical arbitrage. For the construction of mean reversion spreads, we establish an empirical reversion time metric and optimize asset coefficients by minimizing this empirical mean reversion time. In the trading phase, we employ a reinforcement learning framework to identify the optimal mean reversion strategy. Diverging from traditional mean reversion strategies that primarily focus on price deviations from a long-term mean, our methodology creatively constructs the state space to encapsulate the recent trends in price movements. Additionally, the reward function is carefully tailored to reflect the unique characteristics of mean reversion trading.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 4.5/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper employs advanced stochastic calculus (Ornstein-Uhlenbeck processes) and reinforcement learning algorithms, indicating high mathematical density. While it tests on real-world US stock data, the reliance on simulated data for algorithm validation and the focus on theoretical framework over detailed implementation specifics (like transaction costs) places it more in the research lab than ready for the trading desk.

Research Flowchart

  flowchart TD
  A["Research Goal: Model-Free<br>Statistical Arbitrage"] --> B["Inputs: Historical<br>Stock Price Data"]
  B --> C["Spread Construction:<br>Minimize Empirical Mean Reversion Time"]
  C --> D["Trading Framework:<br>Reinforcement Learning RL"]
  D --> E["State Space:<br>Recent Price Trends"]
  D --> F["Reward Function:<br>Tailored for Mean Reversion"]
  E & F --> G["Key Findings:<br>Optimal Trading Strategy &<br>Model-Free ARBITRAGE"]

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 model‑free reinforcement‑learning framework for statistical arbitrage that uses an empirical mean‑reversion time metric to construct spreads and a tailored reward function to trade them.

Source sentence
we introduce an innovative model-free and reinforcement learning based framework for statistical arbitrage.

Assumes 3

  • A constant per‑trade transaction cost c is used in the reward formulation. execution
    Source sentence
    The term c represents the transaction cost per trade.
  • Daily closing prices are an accurate representation of market prices for trading decisions. data
    Source sentence
    Data on the daily closing prices for these stocks was collected over the period from January 1, 2022, to December 31, 2023.
  • Parameters of the Ornstein‑Uhlenbeck process remain constant during the trading phase (criticised by authors). model
    Source sentence
    These estimations depend on historical data from the formation period, with an underlying assumption that parameters remain constant in the following trading phase

Data

  • US stock pair daily closing prices · Yahoo! Finance API market-prices public
    2022-01-01 to 2023-12-31 · Pairs such as MSFT‑GOOGL, CVS‑JNJ, etc. across multiple sectors
    Source sentence
    Data on the daily closing prices for these stocks was collected over the period from January 1, 2022, to December 31, 2023. The data was sourced from the Yahoo! Finance API1.

Recognised sources: Yahoo Finance

Costs

Unclear The classifier was not confident.

Finds 1

  • The reinforcement‑learning method significantly outperforms traditional benchmarks, achieving higher daily Sharpe ratios and larger cumulative returns (e.g., 0.1250 vs 0.0955 daily SR and 37.76% vs 11.44% cumulative PnL for the MSFT‑GOOGL pair). performance
    Source sentence
    Our experimental results demonstrate that the proposed reinforcement learning approach significantly outperforms traditional benchmarks in terms of daily Sharpe Ratio and cumulative returns

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 8 related papers). Coverage is partial, so this is neither replication nor refutation.

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 (34)

  1. Pairs trading: Performance of a relative-value arbitrage rule (2006)
  2. Statistical arbitrage in the US equities market (2010)
  3. Dynamic Modeling of Mean Reverting Spreads for Statistical Arbitrage (2011)
  4. Optimal Mean Reversion Trading: Mathematical Analysis and Practical Applications (2016)
  5. Optimal mean reversion trading with transaction costs and stop-loss exit (2015)
  6. Arbitrage in Stock Index Futures (1990)
  7. Optimal arbitrage strategies on stock index futures under position limits (2011)
  8. A profit model for spread trading with an application to energy futures (2010)
and 26 more
  1. Dynamic modeling of mean-reverting spreads for statistical arbitrage (2011)
  2. Identifying small mean-reverting portfolios (2011)
  3. Sparse Mean-Reverting Portfolios via Penalized Likelihood Optimization (2020)
  4. Constructing cointegrated cryptocurrency portfolios for statistical arbitrage (2019)
  5. Are pairs trading profits robust to trading costs? (2012)
  6. Pairs Trading: quantitative methods and analysis (2004)
  7. Co-integration and error correction: representation, estimation, and testing (1987)
  8. Trading in the presence of cointegration (2012)
  9. Pairs trading and selection methods: is cointegration superior? (2015)
  10. Pairs trading (2005)
  11. A new approach to modeling and estimation for pairs trading (2006)
  12. On the efficacy of optimized exit rule for mean reversion trading (2020)
  13. Sparse mean-reverting portfolios via penalized likelihood optimization (2020)
  14. Pairs trading: A copula approach (2013)
  15. Pairs trading with copulas (2016)
  16. Deep Learning Statistical Arbitrage (2021)
  17. Enhancing a pairs trading strategy with the application of machine learning (2020)
  18. Pairs trading on different portfolios based on machine learning (2021)
  19. Optimal liquidation of a pairs trade (2011)
  20. Stochastic optimization methods for buying-low-and-selling-high strategies (2009)
  21. Differential equations driven by rough paths (2007)
  22. A primer on the signature method in machine learning (2016) — link
  23. Optimal execution with rough path signatures (2020)
  24. Optimal stopping with signatures (2021) — link
  25. Uniqueness for the signature of a path of bounded variation and the reduced path group (2010)
  26. Important extrema of time series (2007)

0 of 34 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.