One of the key algorithms in the model architecture used in this paper is Transformer Encoder (TE). TE is a part of sequence representation extraction models (SREM), which till recently was dominated by Long Short-Term Memory (LSTM). Unlike RNNs, and LSTM in particular, TE processes the entire input all at once, without modeling sequential dependencies in the time series. Given the low auto-correlation structure of financial data, e.g. stock returns, TE is perhaps the most suitable for finance applications among all other currently available ML approaches.
ASSET CLASS: stocks | REGION: United States | FREQUENCY:
Yearly | MARKET: equities | KEYWORD: Factor Allocation, Reinforcement, Learning
I. STRATEGY IN A NUTSHELL
This strategy uses Transformer Encoder and Reinforcement Learning to manage long-horizon multifactor portfolios. By analyzing multiple risk factors, asset characteristics, and rebalancing frequencies, it improves risk-adjusted returns for patient investors.
II. ECONOMIC RATIONALE
The strategy works because long-term, patient investors can optimize factor exposures over extended horizons. Reinforcement Learning accounts for volatility, liquidity, and asset turnover, enabling superior portfolio performance compared to short-horizon trading.
III. SOURCE PAPER
Multi-(Horizon) Factor Investing with AI [Click to Open PDF]
Ruslan Goyenko, McGill University – Desautels Faculty of Management; Chengyu Zhang, McGill University – Desautels Faculty of Management
<Abstract>
We provide a novel approach for multi-factor investing with big data by a multi-horizon investor who takes into consideration long-term versus short-term volatility, liquidity and trading costs trade offs while maximizing expected portfolio returns. Reinforcement learning (RL), which is generally used to solve problems with long- versus short-term reward trade-offs, allows explicitly incorporating long, ten-year investment horizon considerations during training. In out-of-sample, testing period we are the first to show the importance of investment horizon effect for portfolio performance. First, RL portfolio of long term investors with annual rebalancing performs competitively vis-à-vis their short-term peers with monthly rebalancing, and outperforms the latter due to lower portfolio rebalancing needs, turnover and trading costs. Second, when both, short and long-term investors are allowed to rebalance monthly, long-horizon portfolio outperforms by being more patient, with more strategic factor timing and turnover strategies spread over multiple months. Short horizon portfolio is less patient, has higher volatility and almost twice higher monthly turnover. Importantly, we identify different fundamental economic signals determining success of long vs. short-term strategies.
IV. BACKTEST PERFORMANCE
| Annualised Return | 7.37% |
| Volatility | 2.73% |
| Beta | N/A |
| Sharpe Ratio | 2.7 |
| Sortino Ratio | N/A |
| Maximum Drawdown | N/A |
| Win Rate | 71% |
