Is This A Regime Change Or Just A Bad Week?

Every strategy has bad weeks. A mean-reversion book will occasionally hit five losing days in a row purely from noise, then bounce right back. But sometimes a bad week is the first sign that the world genuinely changed — correlations that used to hold have broken, or a factor that worked for years has stopped working — and no amount of patience will bring the old pattern back. The problem is that both situations look identical while you're inside them. A drawdown is a drawdown; it doesn't come labeled "temporary" or "permanent."

This matters because the two situations call for opposite responses. If it's noise, the right move is to sit still, maybe even add to the position at better prices, and let mean reversion do its job. If it's a real regime change, sitting still is how you turn a manageable loss into a career-ending one. Reacting to noise as if it were a regime change is just as costly in the other direction — you cut a strategy that was about to recover, lock in the loss, and miss the reversal.

What actually distinguishes the two

There's no single statistic that settles this cleanly, but a few things shift the odds. First, look at whether the loss is coming from the strategy's known risk factor behaving badly (which is expected, if painful) or from something structural — the correlations the strategy relies on breaking down, the spread it trades no longer mean-reverting at all, or volume drying up so it can't even execute. The former is a bad week; the latter is a warning sign. Second, check whether the move lines up with an identifiable, plausibly temporary cause — a single macro print, a single large seller — versus a slow grind with no obvious trigger, which is more consistent with a genuine shift in the underlying economics. Third, compare the current drawdown's shape to the strategy's own history: if five-day losses this size have happened ten times before and always recovered, that's real evidence about what "normal" looks like for this book specifically, not a guess.

None of these tests is conclusive on its own, and that's the honest answer: you're always making a probabilistic call with incomplete information, not proving anything.

What this means in practice

Desks handle this by pre-committing to rules before the pressure of a live loss clouds judgment — a drawdown limit that triggers a mandatory review rather than an automatic exit, so a human (or a model) actively re-examines the strategy's assumptions instead of either freezing or panic-selling. The review asks a narrow question: has anything about why this strategy should work actually changed, or has it just had a run of bad luck within its known range? Waiting for certainty before acting is not free either — by the time a regime change is obvious to everyone, most of the damage is already done, and the trade to get out has gotten more expensive.