Evaluating Machine-Generated Trading Signals with Formulaic Alphas
Summary
The article introduces formulaic alphas as algebraic representations of trading signals. It argues that expressing signals as formulas can make large collections of machine-generated ideas easier to inspect, combine, and audit. This is presented as a response to automated searches that produce many weak or transient signals, while making manual review impractical and increasing the risk of false discoveries from repeated trials.
It explains the building blocks of an alpha: functions and operators applied to market data, including price, volume, returns, and market capitalization. A simple example uses the relationship between a stock’s intraday change and range to produce a directional score for subsequent trading. The article also outlines common signal families such as momentum and mean reversion, and describes formula expressions as executable models. Its scope is primarily conceptual, with a partial MetaTrader implementation; the supplied text does not provide a complete evaluation procedure or performance evidence. Formulaic form may improve transparency, but it does not establish that a signal has a durable economic edge or avoid data-mining bias.
Key ideas
- Formulaic alphas encode trading signals as algebraic expressions over market data.
- A simple example maps an intraday price move relative to its range into a directional score.
- Operators can capture cross-sectional rankings, time-series behavior, decay, and group neutralization.
- Automated searches create many candidate signals, making manual inspection difficult and increasing false discovery risk.
- Readable formulas can support auditing and combination, but do not prove predictive value.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.