Deciding To Switch A Strategy Off

Every strategy with a real edge still loses money some of the time — that's what it means for returns to be uncertain rather than guaranteed. The hard problem isn't noticing a losing stretch; it's telling whether a given losing stretch is the kind the strategy's own track record says to expect, or a sign that whatever made the strategy work has actually changed or disappeared. Turning a strategy off too early, in response to a drawdown well within its historical range, throws away expected future profits for no real reason. Leaving it on too long after the underlying edge has genuinely eroded burns capital on something that no longer has any reason to work.

The way to make this decision without relying on gut feel in the moment is to set the criteria in advance, before a drawdown is underway and emotions are involved. Two kinds of evidence matter, and they're different. The first is purely statistical: is the current drawdown larger, or has it lasted longer, than what the strategy's backtest and live history say should happen a meaningful fraction of the time — a way of asking "is this still consistent with the strategy just being unlucky." The second is structural: has something about the market, the strategy's opportunity set, or the competitive landscape actually changed in a way that explains why the edge might be gone — new competitors trading the same signal, a regulatory change, a market structure shift that removed the inefficiency the strategy exploited.

Worked example

A strategy has a backtested worst historical drawdown of 12% and average drawdowns that last about six weeks. It is currently down 9% over four weeks — uncomfortable, but inside the range the strategy's own history says to expect periodically. On the statistical criterion alone, this doesn't justify turning it off. But the desk also learns that a specific market structure change — a new exchange rule that reduced the latency advantage the strategy depended on — took effect three weeks ago, right around when the drawdown started. That's a structural reason, independent of the drawdown's size, to suspect the edge itself has degraded rather than just having a normal bad stretch, and it tips the decision toward turning the strategy off or resizing it sharply, even though the statistical evidence alone wouldn't have been enough.

Without both kinds of evidence written into a pre-agreed decision framework, this call tends to get made on vibes — after a string of losses feels bad enough, or after a string of gains feels good enough to ignore mounting warning signs.