Paper: arXiv 2611.23723

Authors: Hui Chen, Haoyu Liu, Ke Wu, Dexin Zhou

Abstract

We study whether visually similar price histories predict similar future returns, using a vision model to retrieve each stock’s most similar charts from three decades of U.S. price histories. A representation pretrained only on natural images already retrieves neighbors whose realized returns predict the cross-section of stock returns. Aligning the representation with subsequent returns more than doubles the strategy’s Sharpe ratio and outperforms a convolutional neural network trained directly on chart images. The predictability is concentrated among stocks with greater limits to arbitrage. Returns and order imbalances around portfolio formation indicate that the signal identifies temporary selling pressure that partially reverts. A sparse autoencoder traces the return-aligned signal to recurring visual states that determine which historical charts are retrieved. Our approach traces each forecast to dated historical charts, making it more transparent than direct image-to-return models and linking rule-based technical analysis with machine-learning chart reading.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 6.0/10 (capped from 9.0: reported Sharpe ratio of 6.37 is implausibly high for daily or lower-frequency data; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 9.0)

Why this score: The paper demonstrates strong empirical rigor through extensive backtesting, robust statistical methodologies, and detailed ablation tests. Its novelty lies in the unique application of visual retrieval with transfer learning and sparse autoencoders to stock return prediction, offering a transparent alternative to direct image-to-return models.

Research Flowchart

  flowchart TD
    A["Research Goal: Do visually similar price histories predict similar future returns?"] --> B{"Methodology: Historical Visual Retrieval"};
    B --> C["Data/Inputs: 3 Decades of U.S. Price Histories (Chart Images)"];
    C --> D{"Computational Process 1: Vision Model (Natural Image Pretrained)"};
    D --> E{"Computational Process 2: Align Representation with Subsequent Returns (Sparse Autoencoder)"};
    E --> F["Key Findings: Predictable Cross-Section of Stock Returns, Concentrated in High 'Limits to Arbitrage' Stocks, Identifies Temporary Selling Pressure"];
    F --> G["Outcome: Transparent forecast linked to dated historical charts, connecting technical analysis & ML chart reading"];

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 historical visual retrieval framework that forecasts a stock’s return from the realized returns of its most similar historical price‑chart images.

Source sentence
We propose a historical visual retrieval framework that forecasts a stock’s return from the realized returns of its most similar historical charts.

Assumes 2

  • The method assumes that for each stock‑date observation the full 20‑day price history and the subsequent five‑day return are available to construct charts and retrieve neighbors. data
    Source sentence
    We require both the complete price history needed to draw a chart and the complete subsequent return needed for its later use as a neighbor.
  • Trading is assumed to be feasible on a weekly basis (queries formed on the last trading day of each week) with the ability to rebalance portfolios at that frequency. execution
    Source sentence
    [Figure 1 Around Here] The historical library is daily, whereas queries are formed on the last trading day of each week. A daily library gives each weekly query a much richer set of dated historical paths than a weekly library would.

Data

Backtest sample: 2001 – 2025 · U.S. stocks · equities · weekly

  • CRSP daily stock file · CRSP market-prices public
    1993-2025 · U.S. common stocks listed on NYSE, AMEX, Nasdaq (excluding ADRs, non‑U.S. firms, abnormal
    Source sentence
    2 Data and Methodology 2.1 Sample and Price-Chart Construction We use daily CRSP data for common stocks listed on the NYSE, AMEX, and Nasdaq, excluding ADRs, non-U.S.-incorporated firms, and observations with abnormal trading status. The sample runs from 1993 to 2025 because daily opening, high, and low prices first become available in June 1992.

Recognised sources: CRSP WRDS

Costs

Modeled Costs are part of the reported headline results.

Costs are charged at 10 basis points per side for stocks above the NYSE 80th market‑cap percentile and 20 basis points otherwise.

Source sentence
Transaction costs are charged on the full absolute weight changes at 10 basis points for stocks above the contemporaneous NYSE 80th market-capitalization percentile and 20 basis points otherwise.

Finds 5

Reported performance (FT-DINO equal-weighted high-minus-low portfolio): Sharpe 6.37 (benchmark 2.81) · annual return 85.0%

  • Using the pretrained DINOv2‑S encoder, an equal‑weighted high‑minus‑low portfolio based on the average outcomes of the 1,000 nearest historical charts earns 26% per year with an annualized Sharpe ratio of 2.81 over 2001–2025. performance
    Source sentence
    Over the 2001–2025 out-of-sample period, an equal-weighted high-minus-low portfolio formed on the average outcomes of the 1,000 nearest historical charts earns 26% per year with an annualized Sharpe ratio of 2.81.
  • After fine‑tuning the encoder on return ranks (FT‑DINO), the same strategy earns 85% per year with a Sharpe ratio of 6.37, more than double the pretrained benchmark. performance
    Source sentence
    With 1,000 neighbors, the FT-DINO equal-weighted high-minus-low portfolio earns 85% per year, with an annualized alpha of 81% and a Sharpe ratio of 6.37, 2.3 times the pretrained benchmark.
  • Under capped value weighting, FT‑DINO earns 28% per year with a Sharpe of 2.38, compared with 0.87 for the pretrained DINO representation. performance
    Source sentence
    Under capped value weighting, which winsorizes market capitalization at the NYSE 80th percentile (Jensen et al., 2023), the portfolio earns 28% per year with a Sharpe ratio of 2.38, compared with 0.87 for pretrained DINO.
  • FT‑DINO outperforms the 20‑day image‑based CNN of Jiang et al. (2023); under equal weighting the CNN’s Sharpe is 4.46 versus FT‑DINO’s 6.37 (a 43% higher Sharpe). performance figure from elsewhere in the paper
    Source sentence
    Under equal weighting, the CNN high-minus-low portfolio earns 61% per year with a Sharpe ratio of 4.46, compared with 85% and 6.37 for FT-DINO, a 43% higher Sharpe ratio.
  • Even after controlling for pretrained DINO, the CNN, and standard factors, FT‑DINO retains a statistically significant annualized alpha of 41% (t‑stat 16.53) under equal weighting, indicating the two signals are not redundant. statistical
    Source sentence
    After controlling for pretrained DINO, the CNN, and standard factors, FT-DINO retains an annualized alpha of 41% (t-statistic of 16.53) under equal weighting and 15% (t-statistic of 8.50) under capped value weighting, about half of its raw return in both cases.2 The two approaches also read the chart differently: the CNN concentrates on the last few trading days,

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 15 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-19 · same card as JSON · A language model chose these items: check the source sentence before relying on one.