Paper: arXiv 2412.00896

Authors: Weizhe Ren, Yichen Qin, Yang Li

Abstract

Traditional genetic programming (GP) often struggles in stock alpha factor discovery due to its vast search space, overwhelming computational burden, and sporadic effective alphas. We find that GP performs better when focusing on promising regions rather than random searching. This paper proposes a new GP framework with carefully chosen initialization and structural constraints to enhance search performance and improve the interpretability of the alpha factors. This approach is motivated by and mimics the alpha searching practice and aims to boost the efficiency of such a process. Analysis of 2020-2024 Chinese stock market data shows that our method yields superior out-of-sample prediction results and higher portfolio returns than the benchmark.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs only partially modelled; see penalty flags)
  • Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 8.0)

Why this score: The paper is primarily methodological, focusing on algorithmic improvements (tree structures, initialization) with minimal advanced mathematics beyond basic statistics. It is highly empirical, evidenced by backtests on real Chinese stock market data from 2020-2024, specific performance metrics like IC and portfolio returns, and a clear comparison to benchmarks.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Efficient Alpha Factor Discovery<br>via Genetic Programming"] --> B{"Data & Inputs"}
  B --> B1["Stock Market Data<br>(2020-2024 China)"]
  B --> B2["Genetic Programming<br>Framework"]
  B --> B3["Structural Constraints<br>& Warm Start"]
  
  B --> C["Methodology: GP Initialization"]
  C --> D{"Optimization Process"}
  D --> D1["Focused Search<br>Promising Regions"]
  D --> D2["Structural Constraints<br>Reduce Search Space"]
  D --> D3["Interpretability<br>Enhancement"]
  
  D --> E["Outcomes & Findings"]
  E --> E1["Superior Out-of-Sample<br>Prediction Results"]
  E --> E2["Higher Portfolio Returns<br>vs Benchmark"]
  
  style A fill:#e1f5e1
  style E fill:#fff2cc

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 warm‑start genetic programming framework that uses a carefully chosen initial individual and structural constraints to focus the search on promising alpha structures.

Source sentence
Based on the two hypotheses outlined above, this paper proposes a framework for mining stock selection factors using genetic programming with given initialization and structural constraints.

Assumes 5

  • An alpha with an IC greater than 0.03 is considered effective. statistical
    Source sentence
    In this experiment, we assume an alpha with an IC greater than 0.03 is considered effective.
  • Risk‑free rate is set to zero in Sharpe ratio calculation. model
    Source sentence
    Retf is the risk‑free rate, which is set to 0 in this paper
  • Holding period for evaluating alphas is 5 trading days. execution
    Source sentence
    We set a 5‑day holding period to evaluate the alpha’s ability to predict stock returns.
  • Stocks cannot be bought if hitting the daily limit‑up and cannot be sold if hitting the limit‑down or being suspended. execution
    Source sentence
    stocks cannot be bought if hitting the daily limit‑up, and cannot be sold if hitting the limit‑down or being suspended.
Show 1 more
  • If a stock is flagged with an ST or *ST warning, it is excluded from the portfolio. execution
    Source sentence
    if a stock is flagged with an ST or *ST warning, we will no longer hold that stock

Data

Backtest sample: 2020 – 2024 · Chinese equities · equities · daily

  • A‑shares data of the Chinese stock market market-prices unknown
    2020.01‑2021.12 (mining) and 2022.01‑202 · Chinese A‑shares
    Source sentence
    We use the full A‑shares data of the Chinese stock market for alpha discovery, with the alpha mining period spanning from 2020.01 to 2021.12. The testing period is set from 2022.01 to 2024.10.

Costs

Partial Costs appear only as a robustness check, a flat fee without spread or impact, or outside the headline numbers.

Finds 5

Reported performance (Warm Start Genetic Programming (WS) alpha framework): Sharpe 1.06 · annual return 56.4%

  • Constraining the search to an effective structure raises the proportion of alphas with IC > 0.03 to over 13%, more than tripling the density from the unconstrained space. performance
    Source sentence
    The proportion of alphas with IC > 0.03 exceeded 13%, more than tripling the density observed in the unconstrained space.
  • Warm‑Start GP produces alphas with lower average absolute correlation (0.60) compared to traditional GP (0.87). statistical
    Source sentence
    a. GP: average absolute value 0.87 b. Warm Start GP: average absolute value 0.60
  • Portfolios built from Warm‑Start GP alphas achieve annualized returns of 48.4% (size = 10) with a Sharpe ratio of 0.937, outperforming both traditional GP and the original Alpha101 alphas. performance
    Source sentence
    WS LR 0.484 0.937 0.564 1.059 0.534 0.959
  • The proposed framework’s alphas generate annualized returns exceeding 50% for holding sizes of 30 and 100 stocks, with Sharpe ratios above 1.0 for size = 30. performance
    Source sentence
    The annualized return exceeds 50% under both 30 and 100 holding sizes, with the Sharpe ratio exceeding 1.0 under the 30 holding size.
  • Out‑of‑sample prediction results and portfolio returns of the proposed framework are superior to benchmark methods on Chinese stock data from 2020‑2024. performance
    Source sentence
    Our analysis of 2020‑2024 Chinese stock market data shows that the proposed framework yields superior out‑of‑sample prediction results and higher portfolio returns than the benchmark.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 31 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive (2 by independent authors): From Deep Learning to LLMs: A survey of AI in Quantitative … (uses its method or data)

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 (26)

  1. Capital asset prices: A theory of market equilibrium under conditions of risk (1964)
  2. Common risk factors in the returns on stocks and bonds (1993)
  3. A five-factor asset pricing model (2015)
  4. Inside the Black Box: A Simple Guide to Quantitative and High-Frequency Trading (2013)
  5. Applied attention-based LSTM neural networks in stock prediction (2018)
  6. Genetic Programming for Quantitative Stock Selection (2009)
  7. Temporal relational ranking for stock prediction (2019)
  8. Alphaevolve: A learning framework to discover novel alphas in quantitative investment (2021)
and 18 more
  1. Research on symbolic regression based on genetic programming (2006)
  2. Genetic Programming: On the Programming of Computers by Means of Natural Selection (1992)
  3. Using genetic algorithms to find technical trading rules (1999)
  4. 101 formulaic alphas (2016)
  5. A Multi-Factor Stock Selection System Based on Short-Term Price and Volume Characteristics (2017)
  6. Stock Selection Factor Mining Based on Genetic Programming (2019)
  7. GPU Acceleration for Genetic Programming Factor Mining (2024)
  8. Genetic Programming Deconstruction and Investment Insights (2023)
  9. Finding Unique Alphas: Genetic Algorithms and Genetic Programming (2016)
  10. Re-examining Stock Selection Factor Mining Based on Genetic Programming (2019)
  11. Open Issues in Genetic Programming (2010)
  12. Applications of Genetic Programming to Finance and Economics: Past, Present, Future (2020)
  13. Constrained Genetic Programming to Minimize Overfitting in Stock Selection (2008)
  14. An In-Depth Investigation of Genetic Programming and Nine Other Machine Learning Algorithms in a Financial Forecasting Problem (2022)
  15. An Overview of Methods Maintaining Diversity in Genetic Algorithms (2012)
  16. Autoalpha: An Efficient Hierarchical Evolutionary Algorithm for Mining Alpha Factors in Quantitative Investment (2020) — link
  17. Genetic Algorithm Based Quantitative Factors Construction (2022)
  18. An Analysis of the Causes of Code Growth in Genetic Programming (2002)

0 of 26 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.