Max Drawdown

A drawdown is any decline from a running equity peak. The maximum drawdown (MDD) is the deepest one in the tested period:

drawdown(t) = (peak so far − equity(t)) / peak so far
MDD         = max over t of drawdown(t)

An account that grows to $12,000, sinks to $9,000, then recovers, took a (12,000 − 9,000) / 12,000 = 25% max drawdown, regardless of where it ended up.

Depth and recovery

Recovery is asymmetric: a 25% decline needs a 33.3% gain from the trough to return to the old peak, while a 50% decline needs a 100% gain. Depth is only one dimension. Drawdown duration measures how long equity stays below its prior high, and the frequency of the equity observations affects what the calculation can see.

Reading it in a backtest

State whether the curve is marked to market throughout each trade or updated only when trades close. Closed-trade equity can hide a large open loss. Compare depth with duration, return, exposure, and leverage; the Calmar ratio, for example, divides annualized return by maximum drawdown.

A common misconception is that historical maximum drawdown is a limit on future loss. It is only the largest decline observed in that sample under that model. The next path can be worse. Monte Carlo simulation can show how trade order changes path risk, but its range is still limited by the trades and assumptions supplied to it.

On AlphaProve

AlphaProve's tearsheet pairs the equity curve with an underwater chart and drawdown duration. Its simulated portfolio can also halt new entries after a configured max_drawdown_pct; that rule changes the tested strategy and does not guarantee a live loss ceiling. The stop-loss experiment, nine ways to place a stop, is a concrete example of why return and drawdown need to be read together.