Market regime analysis for automated trading strategies
A strategy can be profitable overall while depending heavily on one type of market. Regime analysis asks whether the edge survives when the market's behavior changes.
What is a market regime?
A market regime is a repeatable state used to describe market behavior. Common examples include trending versus ranging conditions, high versus low volatility, expansion versus compression, or directional versus rotational structure. There is no single correct regime definition; the useful definition is one that can be measured consistently and tested without looking ahead.
Why regime sensitivity matters
Aggregate backtest statistics can hide dependence. A strategy may produce nearly all of its profits during strong directional markets while slowly losing during ranges. Another system may thrive in mean reversion but fail during volatility expansion. Knowing that relationship changes the research question from “Does the strategy work?” to “Under what observable conditions does its behavior change?”
How to test regime sensitivity
Define the state before analyzing outcomes
Use measurements available at or before the trade decision. Avoid regime labels that depend on knowing what happened afterward. Otherwise the analysis becomes a look-ahead classification exercise rather than something that could be used in live research.
Compare more than net profit
For each regime, compare trade count, expectancy, win rate, average win and loss, MAE, MFE, drawdown contribution and holding time. A regime may have lower win rate but still produce valuable asymmetric winners.
Check stability across time
A relationship found in one month or contract can disappear in another. Split the data chronologically and test whether the same regime relationship appears in later samples.
Do not automatically block the worst regime
A weak regime can still contain important trades. If it produces both large losses and occasional large winners, simply removing it may damage the strategy. Alternatives worth testing include reduced risk, different exit logic, stronger entry confirmation, or a separate management policy.
Regime analysis should produce hypotheses
A useful result is specific: “The strategy's short trades show materially worse follow-through during low-volatility rotational conditions, while long performance is comparatively stable.” That can lead to a controlled test. Generic labels like “bad market” cannot.
EdgeQuery is a research platform for automated traders that analyzes trade history alongside market context. The objective is to diagnose why a system works, where it struggles, and what may be worth testing next—not to provide investment advice or promise future performance.
Join the founding beta waitlist