Resampled Path Analytics

A single backtest equity curve shows only the one trade order that happened to occur. The trades themselves - their sizes, their win/loss mix - are fixed, but the ORDER they arrived in is one draw out of many possible orders, and a strategy's apparent smoothness (or roughness) can owe more to that one lucky (or unlucky) ordering than to any genuine property of the trades. RpaResample() takes a historical trade P&L series and generates many independent synthetic paths by sampling trades WITH replacement, in a fresh random order per path, walking each one out to a fixed number of trades from a common starting equity. Across all the simulated paths, it reports two things: a percentile envelope (p5/p25/p50/p75/p95) of equity at every step, showing how wide the plausible range of outcomes really is at each point along the way, and a ruin probability - the fraction of simulated paths whose equity ever touched or fell below a caller-specified ruin level at any point, not just at the end. This is a standard trade-sequence bootstrap, not a market-return bootstrap - it reshuffles the ALREADY-REALISED trade outcomes rather than generating new hypothetical trades, so it answers "how much did trade order matter to the curve I got" rather than "what other trades might this strategy have produced under different market conditions". Notes and limitations - The output percentiles[] array is a FLAT 1D array, not a genuine 2D array - MQL4 cannot resize both dimensions of a real 2D array at runtime (only the leftmost dimension is resizable), so this uses row-major manual indexing (step * RPA_BAND_COUNT + bandIndex) instead. Always read it through RpaPercentileAt() rather than indexing the array directly. - This is a trade-sequence bootstrap, not a market bootstrap - see the full description above. It cannot surface a loss pattern that never occurred in the source trades at all, only different ORDERINGS of the trades that did occur. - Ruin is checked at EVERY step along a path, not only at the final step - a path that dips to the ruin level midway and later recovers above it still counts as ruined, matching how a real account would actually be stopped out rather than allowed to keep trading through a margin call. - MathRand() returns a value in 0..32767; sampling uses `MathRand() % tradeCount`, which is a standard, adequate bootstrap sampler for the modest trade counts and path counts this header targets, though it is not a cryptographically uniform sampler (an academic-grade concern, not a practical one here). - Cost is O(pathCount * pathLength) for the simulation and O(pathLength * pathCount * log(pathCount)) for the per-step sort - keep both counts in the low hundreds for interactive use; a few thousand of either will noticeably slow a script run. - seed=0 does not mean "no randomness" - it means "don't reseed", reusing whatever state MathRand()'s internal generator is already in. Pass a genuinely fixed non-zero seed whenever a reproducible result across separate runs is required.

Script · MT4 · v1.0 · Free

Tags: Channels & Envelopes · MT4 Script · MQL4 Source Code

MQL4 & MQL5 scripts · PolyMaria