About this app

A forensic examination of a backtest rather than another backtest. Paste a trade ledger and it runs an execution cost frontier to find the basis points at which the edge dies, reorders the trades four thousand times to see whether the sequence flattered you, bootstraps the expectancy for a confidence interval, reads the record chronologically for out-of-sample decay, strips the best trades to test concentration, removes random trades, and applies the Bailey and Lopez de Prado deflated Sharpe against the number of variants you admit to trying. Where market data is reachable it also compares against buy and hold, splits profit by trend and volatility regime, and runs six hundred random-entry placebo strategies to check the edge beats noise. Look-ahead bias, survivorship and parameter fragility are declared rather than detected — they are not visible in a list of trades — and are deliberately excluded from the score. Every attack that cannot run says so instead of being assumed passed.