Skip to content

Calibration / State contrasts

Compare adjacent state buckets.

Select a sport and adjacent time buckets to inspect the published outcome-frequency difference while keeping both underlying populations in view.

Reading trail

Read first: Read blowout timing before comparing published adjacent state buckets.

Next question: How much independent support remains when observations repeat within a game?

Read the analysis: When a lead lasts (sources regenerated)

PaperSupporting sourceComebacks by deficit and time (sources regenerated)NBA | as_of 2026-09-16PaperSupporting sourceWhen a lead lasts (sources regenerated)Cross-sport | as_of 2026-09-16PaperSupporting sourceWhere the forecast errors live (sources regenerated)Cross-sport | as_of 2026-09-16ModuleSupporting sourceWhy AttributionCross-sport | as_of 2026-09-17InspectorPrerequisiteBlowout timingCross-sport | as_of unrecordedPrerequisite: Read comeback rates by deficit and time remaining before treating a margin as permanent.ModuleNext questionBlowout DynamicsCross-sport | as_of 2026-09-17

These are BETWEEN-BUCKET differences in outcome frequency, not transition frequencies and not paired movements within individual games. Holding the forecast band fixed still does not follow identical games across time buckets.

MLB bucket contrasts
Each row compares published bucket populations from early(inn1-3) to mid(inn4-6). The bar is scaled only within this selected time-pair.

From state: early(inn1-3) / .6-.8

Outcome frequency 65.17% | Forecast observations 2,702

To state: mid(inn4-6) / .6-.8

Outcome frequency 76.64% | Forecast observations 2,449

Difference +11.47 pp | Minimum support 2,449

From state: early(inn1-3) / .2-.4

Outcome frequency 24.01% | Forecast observations 1,033

To state: mid(inn4-6) / .2-.4

Outcome frequency 17.94% | Forecast observations 2,285

Difference -6.07 pp | Minimum support 1,033

From state: early(inn1-3) / .4-.6

Outcome frequency 48.82% | Forecast observations 5,754

To state: mid(inn4-6) / .4-.6

Outcome frequency 50.89% | Forecast observations 2,468

Difference +2.07 pp | Minimum support 2,468

Published contrasts from early(inn1-3) to mid(inn4-6). Both state populations remain visible.
From stateOutcome frequencyForecast observationsTo stateOutcome frequencyForecast observationsDifferenceMinimum support
early(inn1-3) / .6-.865.17%2,702mid(inn4-6) / .6-.876.64%2,449+11.47 pp2,449
early(inn1-3) / .2-.424.01%1,033mid(inn4-6) / .2-.417.94%2,285-6.07 pp1,033
early(inn1-3) / .4-.648.82%5,754mid(inn4-6) / .4-.650.89%2,468+2.07 pp2,468
public/data/showcase/why_attribution.jsonDate not published.n not published

Source fields: public/data/showcase/why_attribution.json -> sports.<sport>.transitions[] -> from.time, from.prob, from.mean_y, from.n, to.time, to.prob, to.mean_y, to.n, winprob_delta, min_support_n.