Paper: arXiv 2406.10695

Authors: Adam Korniejczuk, Robert Ślepaczuk

Abstract

The study seeks to develop an effective strategy based on the novel framework of statistical arbitrage based on graph clustering algorithms. Amalgamation of quantitative and machine learning methods, including the Kelly criterion, and an ensemble of machine learning classifiers have been used to improve risk-adjusted returns and increase immunity to transaction costs over existing approaches. The study seeks to provide an integrated approach to optimal signal detection and risk management. As a part of this approach, innovative ways of optimizing take profit and stop loss functions for daily frequency trading strategies have been proposed and tested. All of the tested approaches outperformed appropriate benchmarks. The best combinations of the techniques and parameters demonstrated significantly better performance metrics than the relevant benchmarks. The results have been obtained under the assumption of realistic transaction costs, but are sensitive to changes in some key parameters.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts like graph clustering algorithms, correlation matrix manipulation, and the Kelly criterion, while providing extensive backtesting on 23 years of S&P500 data with rigorous parameter optimization, cross-validation, and sensitivity analysis.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Develop Effective Statistical<br>Arbitrage Strategy"] --> B["Data & Inputs:<br>US Equities Market Data,<br>Realistic Transaction Costs"]
  B --> C["Key Methodology:<br>Graph Clustering<br>for Pair Selection"]
  C --> D["Computational Process:<br>Ensemble ML Classifiers<br>+ Kelly Criterion"]
  D --> E["Optimization:<br>Dynamic Take Profit<br>& Stop Loss Functions"]
  E --> F["Key Outcomes:<br>Outperformed Benchmarks<br>Risk-Adjusted Returns<br>Realistic Cost Sensitivity"]

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 statistical arbitrage trading strategy that combines graph clustering algorithms with an ensemble of machine learning classifiers, Kelly criterion, and novel time‑variant stop‑loss and take‑profit functions.

Source sentence
The study seeks to develop an effective strategy based on the novel framework of statistical arbitrage based on graph clustering algorithms.

Assumes 6

  • Transaction costs are modeled as a flat 0.05% fee applied at both opening and closing of positions. execution
    Source sentence
    Unless stated otherwise, transaction costs equal to 0.05% have been applied, both at the closing and opening of the position.
  • Fractional shares can be bought and no excess capital is required for short‑position deposits, allowing zero‑cost arbitrage at inception. execution
    Source sentence
    the assumption of the possibility of buying fractional shares is made, as well as the assumption of no excess capital being needed for short positions deposit, thus allowing for zero cost arbitrage at inception.
  • The results are obtained under realistic transaction‑cost assumptions but are sensitive to changes in key parameters. model
    Source sentence
    The results have been obtained under the assumption of realistic transaction costs, but are sensitive to changes in some key parameters.
  • Price data are accurate historical daily prices of S&P 500 constituents downloaded from Yahoo Finance. data
    Source sentence
    The work was based on the price data of historical constituents of the S&P500 index, downloaded from the Yahoo finance database.
Show 2 more
  • Rebalancing occurs on a fixed schedule (every 10th trading day in the modified version). execution
    Source sentence
    the described strategy has been altered, such that the rebalance of the portfolio occurs every 10th trading day, instead of every 3rd one.
  • The backtesting starts with a $1,000 capital base and reinvests profits. execution
    Source sentence
    In each testing instance, the equity curve is constructed by starting with 1000 dollars and reinvesting the profits.

Data

Backtest sample: 2006-03 – 2022-12 · S&P 500 stocks · equities · daily

  • Historical S&P 500 constituents price data · Yahoo Finance market-prices public
    03.2000 to 12.2022 · S&P 500 constituents
    Source sentence
    The work was based on the price data of historical constituents of the S&P500 index, downloaded from the Yahoo finance database.
  • Own implementation backtest dataset benchmark unknown
    03.2000 to 12.2022 · S&P 500 constituents
    Source sentence
    The strategies testing period lasted from 03.2000 to 12.2022.

Recognised sources: Yahoo Finance

Costs

Modeled Costs are part of the reported headline results.

Flat 0.05% fee applied at both opening and closing of each trade.

Source sentence
Unless stated otherwise, transaction costs equal to 0.05% have been applied, both at the closing and opening of the position.

Finds 3

Reported performance (ensemble classifier with Kelly criterion and time‑variant stop‑loss/take‑profit): annual return 49.3% · max drawdown 32.0%

  • The original SPONGEsym‑based strategy achieved annualized returns of 12.2%, Sharpe ratio 1.1 and Sortino ratio 2.01. performance
    Source sentence
    The strategy utilizing SPONGEsym outperformed the strategy using fixed FamaFrench sector-based clusters and significantly outperformed the buy and hold benchmark, achieving annualized returns equal to 12.2%, Sharpe ratio of 1.1 and Sortino ratio of 2.01.
  • Transaction costs incurred are four times larger than the net profit of the strategy under 0.05% cost assumption. performance
    Source sentence
    The amount of transaction costs incurred is four times larger than the net profit from the strategy, as visible in Figure 2.
  • All variants of the classifier‑based strategy outperformed both the original strategy and the benchmark in the out‑of‑sample period. performance
    Source sentence
    all of the variants of the classifier-based strategy outperformed both the ori…

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 30 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors): Forecasting Equity Correlations with Hybrid Transformer … (discusses its findings)

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.