Paper: arXiv 2305.08241

Authors: William H. Press

Abstract

Trade prices of about 1000 New York Stock Exchange-listed stocks are studied at one-minute time resolution over the continuous five year period 2018–2022. For each stock, in dollar-volume-weighted transaction time, the discrepancy from a Brownian-motion martingale is measured on timescales of minutes to several days. The result is well fit by a power-law shot-noise (or Gaussian) process with Hurst exponent 0.465, that is, slightly mean-reverting. As a check, we execute an arbitrage strategy on simulated Hurst-exponent data, and a comparable strategy in backtesting on the actual data, obtaining similar results (annualized returns \(\sim 60\)% if zero transaction costs). Next examining the cross-correlation structure of the \(\sim 1000\) stocks, we find that, counterintuitively, correlations increase with time lag in the range studied. We show that this behavior that can be quantitatively explained if the mean-reverting Hurst component of each stock is uncorrelated, i.e., does not share that stock’s overall correlation with other stocks. Overall, we find that \(\approx 45\)% of a stock’s 1-hour returns variance is explained by its particular correlations to other stocks, but that most of this is simply explained by the movement of all stocks together. Unexpectedly, the fraction of variance explained is greatest when price volatility is high, for example during COVID-19 year 2020. An arbitrage strategy with cross-correlations does significantly better than without (annualized returns \(\sim 100\)% if zero transaction costs). Measured correlations from any single year in 2018–2022 are about equally good in predicting all the other years, indicating that an overall correlation structure is persistent over the whole period.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.0)

Why this score: The paper employs advanced statistical mechanics concepts (fractional Brownian motion, Hurst exponent, shot-noise processes) and requires sophisticated analysis to derive its results, justifying a high math score. It is highly data- and implementation-heavy, using a large, multi-year dataset of ~1000 stocks at minute resolution, conducting backtests on simulated and actual data to quantify arbitrage returns, which warrants a high empirical rigor score.

Research Flowchart

  flowchart TD
  A["Research Goal: Quantify NYSE Price Dynamics<br>& Arbitrage Potential"] --> B["Data: 1000 NYSE Stocks<br>2018-2022, 1-min resolution"]
  
  B --> C{"Analysis Phase 1:<br>Time-Series & Autocorrelation"}
  B --> D{"Analysis Phase 2:<br>Cross-Correlation Structure"}
  
  C --> E["Fit Hurst Exponent H ≈ 0.465<br>(Mean-Reverting)"]
  D --> F["Find Correlations Increase<br>with Time Lag"]
  
  E --> G["Arbitrage Strategy 1:<br>Exploiting Mean Reversion"]
  F --> H["Arbitrage Strategy 2:<br>Exploiting Cross-Correlations"]
  
  G --> I["Outcomes:<br>~60% Annualized Return"]
  H --> J["Outcomes:<br>~100% Annualized Return<br>Variance Explained: 45%<br>Persistent over 2018-2022"]
  
  style A fill:#f9f,stroke:#333,stroke-width:2px
  style I fill:#bbf,stroke:#333,stroke-width:2px
  style J fill:#bbf,stroke:#333,stroke-width:2px

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 an arbitrage strategy that exploits cross‑stock correlations, achieving higher returns than a strategy without such information.

Source sentence
An arbitrage strategy with cross‑correlations does significantly better than without (annualized returns ∼100% if zero transaction costs).

Assumes 6

  • The correlated part of the variance is memoryless (V(τ) ∝ τ). model
    Source sentence
    Next rashly assume that the correlated part of the variance is memoryless, V (τ) ∝τ, so that all of the arbitrageable signal
  • The mean‑reverting Hurst component of each stock is uncorrelated with other stocks. model
    Source sentence
    can be quantitatively explained if the mean‑reverting Hurst component of each stock is uncorrelated, i.e., does not share that stock’s overall correlation with other stocks.
  • Zero transaction costs are assumed for reported headline returns. execution
    Source sentence
    annualized returns ∼60% if zero transaction costs
  • Trades can be executed at the average price of the hour without market impact. execution
    Source sentence
    we assume that our purchases and short sales do not move the market, and that they can be accomplished during a trading hour at the average price of that hour
Show 2 more
  • A dollar turns over ~8000 times per year, requiring round‑trip costs ≲10⁻⁴ for profitability. market
    Source sentence
    Since a given dollar turns over ∼8000 times per year, round‑trip transaction costs ≲10⁻⁴
  • The cross‑correlation structure is persistent across years. statistical
    Source sentence
    Measured correlations from any single year in 2018–2022 are about equally good in predicting all the other years, indicating that an overall correlation structure is persistent over the whole period.

Data

Backtest sample: 2018 – 2022 · NYSE-listed stocks · equities · intraday

  • NYSE one‑minute candles · Finnhub.io market-prices public
    2018–2022 · ≈1000 NYSE‑listed stocks
    Source sentence
    for NYSE‑listed stocks in the years 2018–2022, comprising substantially all such listed stocks.

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
annualized returns ∼60% if zero transaction costs

Finds 5

Reported performance (cross-correlation arbitrage strategy): annual return 100.0%

  • The price discrepancy from a Brownian‑motion martingale follows a power‑law shot‑noise process with Hurst exponent 0.465, indicating slight mean‑reversion. statistical
    Source sentence
    is well fit by a power‑law shot‑noise (or Gaussian) process with Hurst exponent 0.465, that is, slightly mean‑reverting.
  • Cross‑stock correlations increase with time lag over the studied range. statistical
    Source sentence
    we find that, counterintuitively, correlations increase with time lag in the range studied.
  • Approximately 45% of a stock’s 1‑hour return variance is explained by its correlations to other stocks, largely due to market‑wide movements. statistical
    Source sentence
    Overall, we find that ≈45% of a stock’s 1‑hour returns variance is explained by its particular correlations to other stocks, but that most of this is simply explained by the movement of all stocks together.
  • Arbitrage back‑testing on actual data yields about 60% annualized return assuming zero transaction costs. performance
    Source sentence
    obtaining similar results (annualized returns ∼60% if zero transaction costs).
  • An arbitrage strategy that uses cross‑correlations achieves roughly 100% annualized return under zero‑cost assumptions. performance
    Source sentence
    An arbitrage strategy with cross‑correlations does significantly better than without (annualized returns ∼100% if zero transaction costs).

Checked by others

Contradicted by

  • Correlation emergence in two coupled simulated limit order books (2024)
    On the question: How does correlation change as the sampling time interval (Δt) increases?
    Compare the findings

    This paper: Cross‑stock correlations increase with time lag over the studied range.

    That paper: Increasing the sampling time Δt leads to larger drops in correlation.

    A finds correlations rise with longer lags (hours‑years) in NYSE data; B finds larger Δt causes bigger drops in correlation in simulated limit‑order books, a contradiction.

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.