You optimized across many variants. How likely is your best one just curve-fit to noise?Give us the runs — your optimizer's results, or a returns matrix of your variants — and we run the gold-standard data-snooping tests: PBO/CSCV, White's Reality Check, Hansen SPA, and a multiple-testing haircut. We never run a backtest ourselves, and none of this measures profitability.
From your MT5/TradingView optimizer table: take the BEST run and tell us how many you tried. We apply a multiple-testing haircut (Harvey & Liu).
PBO, from the CSCV method (Bailey, Borwein, López de Prado & Zhu, 2015), estimates how often the variant that looks best in-sample underperforms out-of-sample across many symmetric splits of your data. A high PBO means your selection process is fitting noise, not finding a real edge. It is not a measure of profitability.
No. You supply the many runs — either your optimizer's results table (the quick haircut) or a returns matrix of your variants (the gold-standard PBO / Reality Check). We only run the overfitting statistics on what you upload; we never execute a strategy.
The quick mode takes your best run's Sharpe, its trade count, and how many variants you tried, and applies a multiple-testing haircut (Harvey & Liu). The gold-standard mode takes a returns matrix (rows = periods, columns = variants) and runs PBO/CSCV plus White's Reality Check and Hansen SPA.
No. Low PBO and a significant Reality Check mean the SELECTION of the best variant is unlikely to be pure overfitting — it says nothing about whether the strategy will make money. Nothing here is a prediction or trading advice.