Strategy Decommissioning Criteria

Every strategy eventually stops working, either because the market pattern it exploited genuinely fades, or because it was never really there and the original backtest was a false positive that took time to reveal itself in live results. The hard part isn't accepting this in the abstract — it's recognizing it in the moment for a specific strategy, while it's losing money, and deciding whether to turn it off. Left undecided in advance, this decision is exposed to exactly the psychological pressures it shouldn't be: a strategy that made money for years earns trust that makes people reluctant to pull it during a losing stretch that might just be normal variance, while a newer strategy with a much smaller track record might get killed after a much milder loss simply because there's less accumulated confidence in it.

Setting the rule before you need it

Decommissioning criteria are written down while the strategy is performing well, precisely because that's the only time they can be set objectively — a rule agreed on during a drawdown is far more likely to be either too lenient (out of hope the loss reverses) or too harsh (out of panic) than one set in advance. A typical criterion might combine a maximum drawdown relative to the strategy's historical worst case, a maximum length of underperformance relative to its backtested Sharpe ratio, and a statistical check on whether recent returns are still consistent with the distribution the strategy was originally validated against.

A concrete case

A strategy is approved for live trading with a written rule: if its drawdown exceeds one and a half times the worst drawdown seen in the full backtest, or if a rolling twelve-month Sharpe ratio falls persistently below zero, it is automatically flagged for review and reduced in size pending investigation. Eighteen months in, the strategy hits the drawdown threshold during a rough quarter. Because the rule was set in advance, the response is a structured review against a pre-agreed bar, not an emotionally charged argument about whether "this time is different" — and the strategy is either reduced, decommissioned, or, if the review finds a clear and temporary explanable cause, kept running with the reasoning documented.

What this means in practice

Decommissioning criteria don't have to be fully automatic — many firms use them as a trigger for mandatory review rather than an instant kill switch — but they do have to be specific and written down before the strategy needs them, or they provide no real protection against the exact biases they're meant to guard against.