Robustness tests: walk forward, Monte Carlo, parameter sensitivity
Walk forward, Monte Carlo and parameter sensitivity, each with a pass threshold set before the test.
Three tests check whether a backtest result holds up: out-of-sample windows, a different order of the same trades, and small changes to every parameter.
| Test | Parameters | Pass when |
|---|---|---|
| Walk forward | 4 quarterly windows, each out of sample against the previous ones | net above zero in at least 3 of 4 windows |
| Monte Carlo | 1,500 paths, 3-year horizon, trade order resampled from the modeled window | 5th percentile max drawdown within the account limit, blow rate at most 20% |
| Parameter sensitivity | Plus and minus 20% on every numeric parameter in the rule sheet | net above zero in at least 70% of combinations |
| Challenge pass simulation | Rules of your prop firm and account size from the Puravida Edge rules dataset (rules version dated in the report): profit target, drawdown model, daily limit, minimum days, consistency rule; 1,500 Monte Carlo paths of the modeled trades, sizing from the rule sheet | informational, no pass or not pass: share of paths that reach the target without breaking a rule, median days to target, and what ended the other paths |
Challenge pass simulation. We replay the strategy's modeled trades through 1,500 Monte Carlo paths under the exact rules of your prop firm and account size: profit target, drawdown model, daily limit, minimum days and consistency rule. The report shows the share of paths that reach the target without breaking a rule, the median days to target, and what ended the other paths. It is a simulation from modeled trades, not a prediction of your evaluation result.
Validation report template
What a tested trading strategy has to show, section by section.
Questions
Why test beyond the backtest?
A backtest shows one history. Walk forward, Monte Carlo and parameter sensitivity show whether the result depends on one stretch of data, one order of trades or one exact parameter value.
What happens if a test is not passed?
The report marks it as not passed and recommends a rule change or a stop. A strategy that did not pass is not deployed.
Account limits: strategies sized for a prop firm account →