The strategy uses CNNs on stock price images to predict returns, forming weekly rebalanced long-short portfolios by market cap, going long top-decile stocks and short bottom-decile ones.
ASSET CLASS: stocks | REGION: United States | FREQUENCY:
Weekly | MARKET: equities | KEYWORD: Image Recognition
I. STRATEGY IN A NUTSHELL
Trades U.S. stocks (NYSE, AMEX, NASDAQ) using a CNN model trained on 20-day tick images of price data. The model uses convolutional, Leaky ReLU, and max-pooling layers to extract spatial price patterns, then applies a softmax classifier to predict upward movement probabilities. A long-short decile portfolio—long top decile, short bottom—is market-cap weighted and rebalanced weekly.
II. ECONOMIC RATIONALE
The CNN captures short-term trends, reversals, and volatility patterns embedded in price images. By learning subtle spatial relationships traditional models miss, it marginally improves prediction accuracy—yet this small edge produces meaningful alpha in equity markets.
III. SOURCE PAPER
(Re-)Imag(in)ing Price Trends [Click to Open PDF]
Jiang, Jingwen and Kelly, Bryan T. and Xiu, Dacheng, University of Chicago, Yale SOM; AQR Capital Management, LLC; National Bureau of Economic Research (NBER), University of Chicago – Booth School of Business; National Bureau of Economic Research (NBER)
<Abstract>
We reconsider the idea of trend-based predictability using methods that flexibly learn price patterns that are most predictive of future returns, rather than testing hypothesized or pre-specified patterns (e.g., momentum and reversal). Our raw predictor data are images—stock-level price charts—from which we elicit the price patterns that best predict returns using machine learning image analysis methods. The predictive patterns we identify are largely distinct from trend signals commonly analyzed in the literature, give more accurate return predictions, translate into more profitable investment strategies, and are robust to a battery of specification variations. They also appear context-independent: Predictive patterns estimated at short time scales (e.g., daily data) give similarly strong predictions when applied at longer time scales (e.g., monthly), and patterns learned from US stocks predict equally well in international markets.
IV. BACKTEST PERFORMANCE
| Annualised Return | 12.11% |
| Volatility | 6.96% |
| Beta | N/A |
| Sharpe Ratio | 1.74 |
| Sortino Ratio | N/A |
| Maximum Drawdown | N/A |
| Win Rate | N/A |
