Walk-forward vs Monte Carlo simulation is a common confusion. Monte Carlo methods shuffle or resample returns/trades to probe how fragile a summary statistic is to ordering and noise.

When Monte Carlo helps

Quick intuition about variance of Sharpe or drawdown on a fixed strategy
Sensitivity to ordering assumptions

When Monte Carlo does not replace walk-forward

Monte Carlo on historical trades can still embed overfitting from the process that generated those trades. Walk-forward tests whether your fitting process survives time (What is WFA?).

Permutation angle

Related: permutation test trading strategy and permutation p-value backtest (p-value article).

Concrete Monte Carlo use cases

Bootstrap returns to see how wide Sharpe confidence intervals might be on a fixed strategy.
Shuffle trade order to see how sensitive a drawdown estimate is to path ordering (when assumptions allow).

When Monte Carlo misleads

If trades are not exchangeable, shuffling is not a faithful model of real uncertainty. If your strategy has strong microstructure dependence, naive shuffles can understate tail risk.

Pair with walk-forward for deployment decisions

For go-live decisions, prefer time-ordered validation that mirrors your research process. Monte Carlo can complement that story, but it should not replace it (What is WFA?).

Block resampling when returns are autocorrelated

If you bootstrap daily returns, use block bootstrap with a block length that matches your dependence horizon. Independent resampling often understates drawdown risk for clustered volatility.

If your conclusions change materially when block length changes, treat tail statements as fragile.

Sample size: Monte Carlo cannot fix too few trades

Monte Carlo repeats what you give it. If the underlying series has 30 trades, you will get a distribution, but it may be nonsense precision.

Pair simulation with trade-count sanity (How many trades).