How to Evaluate Algorithmic Trading Models with Costs and Risk Metrics

Summary

This article explains how to assess a systematic futures model through test design and several complementary performance measures. It recommends testing across multiple instruments and long historical periods, and including commissions and slippage. The discussion emphasizes that short-interval, high-turnover, arbitrage, and high-frequency strategies are especially sensitive to execution costs, so cost assumptions should reflect plausible live trading conditions.

It defines annualized return, absolute and percentage drawdown, return-to-drawdown ratio, Sharpe ratio, win rate, and profit-to-loss ratio. Examples illustrate why a higher win rate can still produce losses, why two models with similar return-to-drawdown ratios can carry different capital risks, and why Sharpe alone can miss severe losses. The article offers useful evaluation heuristics, but its metrics and examples do not establish future performance. Results depend on the test period, cost assumptions, capital management, and the investor’s tolerance for drawdowns; the suggested thresholds are not universal.

Key ideas

• Test models across a broader range of instruments and market histories to examine robustness.
• Include commissions and realistic slippage, especially for high-turnover strategies.
• Assess returns alongside both the size and proportional rate of drawdowns.
• Use Sharpe ratio as one risk-adjusted measure, while recognizing that it may not capture extreme losses.
• Interpret win rate together with the average scale of wins and losses.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.