The strategy uses LSTM to predict intraday returns of S&P 500 stocks, going long on stocks likely to outperform and short on those underperforming, with intraday rebalancing and equal weighting.
ASSET CLASS: stocks | REGION: United States | FREQUENCY:
Intraday | MARKET: equities | KEYWORD: Machine Learning
I. STRATEGY IN A NUTSHELL
The strategy trades S&P 500 stocks using an LSTM model trained on a 4-year window with 1-year stride. Features include intra-day returns and relative returns, standardized to remove outliers. Stocks are classified by predicted probability of outperforming the cross-sectional median. Long positions target top-performing stocks; short positions target the lowest. Portfolio is intraday rebalanced and equally weighted.
II. ECONOMIC RATIONALE
Machine learning leverages large datasets to detect patterns beyond human intuition, reducing overfitting. By learning from historical intraday returns and multiple features, the model exploits cross-sectional inefficiencies, generating systematic long/short opportunities.
III. SOURCE PAPER
Forecasting directional movements of stock prices for intraday trading using LSTM and random forests [Click to Open PDF]
Pushpendu Ghosh, Department of Computer Science & Information Systems, BITS Pilani K.K. Birla Goa Campus, India; Ariel Neufeld, Division of Mathematical Sciences, Nanyang Technological University, Singapore; Jajati Keshari Sahoo, Department of Mathematics, BITS Pilani K.K. Birla Goa Campus, India
<Abstract>
We employ both random forests and LSTM networks (more precisely CuDNNLSTM) as training methodologies to analyze their effectiveness in forecasting out-of-sample directional movements of constituent stocks of the S&P 500 from January 1993 till December 2018 for intraday trading. We introduce a multi-feature setting consisting not only of the returns with respect to the closing prices, but also with respect to the opening prices and intraday returns. As trading strategy, we use Krauss et al. (2017) and Fischer & Krauss (2018) as benchmark and, on each trading day, buy the 10 stocks with the highest probability and sell short the 10 stocks with the lowest probability to outperform the market in terms of intraday returns – all with equal monetary weight. Our empirical results show that the multi-feature setting provides a daily return, prior to transaction costs, of 0.64% using LSTM networks, and 0.54% using random forests. Hence we outperform the single-feature setting in Fischer & Krauss (2018) and Krauss et al. (2017) consisting only of the daily returns with respect to the closing prices, having corresponding daily returns of 0 .41% and of 0.39% with respect to LSTM and random forests, respectively.
IV. BACKTEST PERFORMANCE
| Annualised Return | 194.32% |
| Volatility | 24.96% |
| Beta | N/A |
| Sharpe Ratio | 4.32 |
| Sortino Ratio | N/A |
| Maximum Drawdown | -39.23% |
| Win Rate | N/A |
