Backtest drawdown is usually shallower than the live one
Posted: Sat Oct 31, 2026 7:16 am
A test shows the worst drawdown over the sample, but the future will probably hold a worse one. The worst case in the history is just one random draw from many possible sequences.
Costs, slippage, and execution problems deepen live drawdowns beyond what clean data suggests. Psychological factors add to it, since a trader living through a losing streak may skip signals or change settings, and each deviation compounds the damage.
A reasonable rule is expecting a drawdown one and a half to two times larger than the backtest's worst. Sizing positions so that such a drawdown is survivable, both financially and emotionally, is a basic form of protection.
Monte Carlo simulations, which reshuffle trade sequences, give a distribution of possible drawdowns. The ninety-fifth percentile figure is a more realistic planning number than the single historical worst.
Planning for a worse outcome than the test shows prevents panic when it arrives.
What multiple of the backtest drawdown is used for planning, and has live trading exceeded it?
Costs, slippage, and execution problems deepen live drawdowns beyond what clean data suggests. Psychological factors add to it, since a trader living through a losing streak may skip signals or change settings, and each deviation compounds the damage.
A reasonable rule is expecting a drawdown one and a half to two times larger than the backtest's worst. Sizing positions so that such a drawdown is survivable, both financially and emotionally, is a basic form of protection.
Monte Carlo simulations, which reshuffle trade sequences, give a distribution of possible drawdowns. The ninety-fifth percentile figure is a more realistic planning number than the single historical worst.
Planning for a worse outcome than the test shows prevents panic when it arrives.
What multiple of the backtest drawdown is used for planning, and has live trading exceeded it?