Historical Scenario Replay

A full-sample backtest can tell you a strategy's average Sharpe ratio was 1.3 over ten years without ever forcing you to confront what happened to it during the two weeks in March 2020 when markets fell 30% and correlations across every asset class went to one. Historical scenario replay is the practice of taking a specific, named historical episode — not a random subperiod, a particular crisis with known dynamics — and re-running the strategy's exact logic through that episode's price and event sequence to see what it actually would have done.

The idea

You pick a scenario with a clear identity: the October 1987 crash, the 2008 financial crisis, the March 2020 COVID crash, a flash-crash day, a specific central bank surprise. You then feed the strategy the real price, volume, and (where relevant) news-event sequence from that exact window, running the strategy's actual entry, exit, sizing, and risk-management logic step by step through it — not a statistical summary of the period, the literal sequence of days.

This differs from an ordinary subperiod backtest in two ways: the scenario is chosen because it's a known type of stress (liquidity dry-up, correlation breakdown, a specific shock), not just an arbitrary date range, and the replay is usually paired with an examination of the mechanism — did the strategy lose money because of the price move itself, because bid-ask spreads widened and execution costs spiked, or because a hedge that normally worked stopped correlating the way it was supposed to.

Worked example

A risk-parity strategy that has run well for eight years is replayed through the exact daily sequence of March 9–23, 2020. Its normal rule scales positions inversely to recent realized volatility; replayed through the crash, realized volatility spikes so fast that the strategy's deleveraging rule can't cut exposure before the worst days already hit, producing a simulated drawdown of 18% in two weeks — far outside anything the full ten-year backtest's average statistics would have suggested. The replay also shows that the bond allocation, normally a diversifier against equity selloffs, fell simultaneously with equities during the most acute days as investors sold everything for cash, breaking the correlation assumption the strategy's risk model depended on. That specific finding — not just "the strategy would have lost money" but "the hedge stopped working because of a cash-scramble correlation spike" — is what a full-sample Sharpe ratio can never surface.

What this means in practice

Historical scenario replay is a standard part of risk sign-off before a strategy goes live with real capital, because regulators, risk committees, and allocators want to know specifically how a strategy behaves in known stress conditions, not just its unconditional statistics. It's also diagnostic: replaying multiple different crisis types (a liquidity crisis, a correlation-breakdown crisis, a slow grinding bear market) tests different failure modes, since a strategy robust to one type of stress can be fragile to another.