nx_fin_pbo.nx

nx_fin_pbo.nx -- Probability of Backtest Overfitting (Bailey/Lopez de Prado CSCV). Completes the overfit- detection suite (deflated-Sharpe was the multiple-testing correction; PBO is the combinatorial one). Given an M-strategy x S-group performance matrix, enumerate every C(S, S/2) train/test split (train = set bits of a popcount-S/2 mask, test = the rest); for each split pick the BEST in-sample strategy and ask whether it lands in the BOTTOM HALF out-of-sample. PBO = fraction of splits where it does. High PBO => your "winner" is likely luck. Pure integer combinatorics -- no transcendental functions, exact + gate-checkable. license_tier: ORIGINAL