A strategy has a bad month and the instinct is immediate: something broke. Sometimes that instinct is right. Often there is not enough evidence yet to know.

One of the hardest problems in systematic trading is distinguishing normal underperformance from a genuine change in the mechanism that created the edge.

Two closely related parts of this decision are Market Regimes Explained: Why a Trading Strategy Can Work in One Market and Fail in Another and Backtest vs Live Trading: 7 Reasons the Same Strategy Can Behave Differently.

Start with the implementation

Before diagnosing market decay, confirm that the system being traded is the system that was tested. Source changes, broker settings, data changes, timezones, execution rules and platform updates can all alter behaviour. A strategy cannot be declared dead because a different implementation performed badly.

Then ask whether the market state changed

Many mechanisms are conditional. A trend strategy can struggle in chop without the trend phenomenon disappearing from markets forever. A mean-reversion strategy can suffer during structural repricing. The question is whether the current conditions fall inside a historically difficult state or whether the behaviour is genuinely outside prior experience.

Compare distributions, not feelings

Historical evidence can provide a range for drawdowns, losing streaks, trade frequency and other behaviours. Live monitoring becomes more informative when current outcomes are compared with those distributions. A bad period that remains within an expected envelope is different from a persistent shift in win rate, payoff, excursion or opportunity frequency.

Watch for several dimensions moving together

A single metric can be noisy. Stronger decay evidence may appear when several things change together: fewer qualifying opportunities, lower favourable excursion, deeper adverse excursion, longer time-to-resolution, worse post-cost expectancy or a shift in the market states where the strategy previously worked.

Do not repair uncertainty with optimization

Changing parameters after a short losing period can make the strategy fit the newest data while destroying its relationship with the original mechanism. The first response to uncertainty should be diagnosis and evidence collection. Modification comes later, if the evidence supports it.

When should you stop?

There is no universal number of losing trades that proves an edge is dead. The stopping rule should be related to the strategy's historical distribution, mechanism, expected failure states and capital risk. The more severe the loss profile, the less patience the system can afford while the diagnosis remains unresolved.