Strategy Post-Mortem Analysis

A strategy post-mortem is a written review, done after a strategy is decommissioned or after a significantly bad drawdown, that tries to answer one question honestly: did this fail because markets changed, because the original research was overfit, because of an operational bug, or because of bad luck within normal variance? Each answer implies a completely different lesson, and skipping the post-mortem means the desk is likely to make the exact same sizing or modelling mistake again under a different strategy's name.

A useful post-mortem separates three timelines side by side: the original backtest performance, the live performance up to the point of failure, and the performance of the same strategy logic re-run on the failure period with hindsight. If live performance tracked the backtest closely and then a specific, identifiable regime shift broke it, that's a market-change story. If live performance never matched the backtest even in calm periods, that points to overfitting or a data-snooping problem in the original research. If a single bad day or software bug drove most of the loss, that's an operational failure, not a strategy failure, and shouldn't taint an otherwise sound idea.

The output is usually a short document: what killed it, whether the signal itself is still believed to be real, and one or two concrete process changes — a new stress test, a stricter capacity check — that should apply to future strategies, not just a eulogy for this one.