Binary Prediction Score Audit
Compute a binary Brier score from groups sharing an identical predicted probability.
Your entries stay in this page’s memory and are not submitted to us. Downloads are generated locally. Reloading or leaving the page clears your entries. Do not enter patient information.
Use fictional or aggregate values only. All defaults are hypothetical; calculation stays in your browser.
Calculation results
Binary Brier score: 0.22.
Observed event proportion: 0.5; mean predicted probability: 0.5; absolute overall gap: 0.
- Row 1: n=10, predicted 0.2, observed 0.3, observed − predicted 0.1.
- Row 2: n=10, predicted 0.8, observed 0.7, observed − predicted -0.1.
Assumptions and limits
Each row must represent cases with exactly the same predicted probability; a bin-average probability cannot recover the original individual Brier score. Score = sum(events(1−p)²+(total−events)p²)/N, on the binary 0–1 scale (lower is better). The overall observed-minus-predicted gap is only calibration in the large, not a complete calibration assessment. Brier score reflects both calibration and discrimination and depends on prevalence; compare models on the same appropriate validation sample. Rows must be disjoint. This supplies no confidence intervals or clinical utility claim. Maximum 100 rows; use aggregate or hypothetical values only.
scikit-learn Brier score definition · Read the practical guide · Download blank worksheet (TXT)