Overview
A stock's price is not set by rational investors applying discounted cash flow analysis, but by emotional individuals who each looks at the same stock from a different angle. Prices move up when demand for a stock exceeds supply, and vice versa. If we can predict how investors will move in and out of a stock, we can predict where that stock is headed.
This strategy uses machine learning to analyze varied types of data to take in the perspective of disparate groups of investors. It combines these perspectives, calculating the likelihood that each stock will outperform the benchmark, and forms portfolios using those most likely to outperform. It then layers in some risk management techniques to protect the portfolio from extreme events.
Sources of Alpha
This strategy combines the following types of data to score each stock.
- Reported fundamentals and ratios
- Forecasted fundamentals
- Technical indicators
- Machine learning models based on price histories
- Quantitative factors
- Options greeks and related measures
- Analyst recommendations
- Insider trades
Portfolio Composition
- Uses proprietary algorithm to assign each stock's weighting
- 100 long positions in individual US equities
- Filters out equities with less than $1 million in average daily trading volume
- Sector constrained to match the market composition
- Biweekly rebalance
- 100% invested
- No leverage
Risk Management
- Stop losses set at levels tailored to each stock's volatility
- Portfolio scaled back to be underleveraged during market turbulence, keeping the portfolio's volatility below a set threshold
Correlations to Popular Quantitative Factor Strategies
- SPY: 0.92
- Book to Price: 0.78
- 12M Momentum: 0.78
- Size: 0.85
- EBIT / EV: 0.84
- Quality: 0.89
*More info available upon request