PBOResult¶

Defined in fynance.research

class PBOResult(pbo, logits, prob_oos_loss, slope, intercept, is_perf, oos_perf)[source]

Result of a pbo run.

Attributes:

pbofloat: Probability of backtest overfitting: the share of IS/OOS splits where the in-sample winner’s out-of-sample relative rank falls at or below the median (logit <= 0).
logitsnumpy.ndarray, shape (n_combinations,): Rank logit ln(r / (1 - r)) of the IS winner’s OOS performance, one per split.
prob_oos_lossfloat: Share of splits where the IS winner’s OOS performance is negative.
slopefloat: OLS slope of OOS performance on IS performance (of the IS winner) across splits — a robust-performing selection procedure has a positive slope; near-zero or negative indicates the IS ranking carries no out-of-sample information.
interceptfloat: OLS intercept of the same regression.
is_perfnumpy.ndarray, shape (n_combinations,): The IS winner’s own in-sample metric value, one per split.
oos_perfnumpy.ndarray, shape (n_combinations,): The IS winner’s out-of-sample metric value, one per split.