Online Learning to Combine Technical Trading Strategies
Summary
The document summarizes research on an adversarial-experts online learning algorithm that adjusts parameters for a zero-cost portfolio of technical trading strategies. The approach combines a collection of strategies into an aggregate portfolio and studies their changing interactions using daily and intraday data from the Johannesburg Stock Exchange. It also describes a hypothesis test for statistical arbitrage, comparisons with a benchmark portfolio algorithm that account for win rates, trading costs, and slippage, and a nonparametric estimate of backtest overfitting probability.
The reported findings are cautious: the daily strategy did not pass the statistical-arbitrage test after costs, while intraday results did not establish statistical arbitrage after costs either. These results do not demonstrate a cost-adjusted arbitrage opportunity in either sampling regime. The summary gives no detailed parameter settings or numerical performance measures, and its conclusions are tied to the studied exchange, data, and strategy set; they should not be generalized without further evidence.
Key ideas
- An adversarial-experts online learning method adjusts parameters for an aggregate of technical strategies.
- The study examines daily and intraday data from the Johannesburg Stock Exchange.
- It evaluates statistical arbitrage after costs and compares performance with a benchmark algorithm.
- Daily results failed the stated arbitrage test after costs, while intraday evidence did not establish arbitrage.
- The study also estimates backtest overfitting probability using a nonparametric method.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.