Findings and revisions

17 findings document how measurements hold up under review. Each exhibit either takes a number apart, deflates its own sample, or publishes a failed result alongside the dated source artifact.

MLB in-game join integrity

Mixed-game tick files and incomplete state rows, and the revision-2 calibration artifacts rebuilt from the segment-clean corpus.

as of 2026-09-16 / mlb

Status: regenerated from the segment-clean corpus

Retractions

The six numbers we took back, with what was wrong and the published correction.

as of 2026-07-23 / all

Source artifact: date not published

Effective sample size

Why repeated within-game rows carry less independent information than their count suggests.

as of 2026-07-25T11:12:42.584390+00:00 / mlb

Source artifact: ess_ledger

Verdict flips

Claim families that changed verdict as more data arrived, in sequence.

as of 2026-07-25T11:13:42.974244+00:00 / all

Source artifact: verdict_flip_anatomy

MLB leaderboards, with the nulls

Descriptive Statcast leaderboards published beside the gate nulls that did not hold.

as of 2026-07-24T21:46:01.294251+00:00 / mlb

Source artifact: mlb_descriptive_leaderboards

Tennis: momentum's grain limit

Point-to-point momentum, game-grain nulls, and measured altitude and travel effects.

as of 2026-07-25T11:13:41.013711+00:00 / tennis

Source artifact: tennis_grain_and_myths

NBA momentum, tested

Structural context findings beside individual carryover shapes that returned null.

as of 2026-07-25T11:13:25.348914+00:00 / nba

Source artifact: nba_momentum_tested

The fourth-quarter shift

Descriptive per-player Q4 versus earlier-quarter rate shifts, with their confounds stated.

as of 2026-07-24T22:39:47.027837+00:00 / nba

Source artifact: nba_q4_shift

When the leaderboard regresses

Published empirical-Bayes movement from raw rates toward their group mean.

as of 2026-07-24T23:02:15.164907+00:00 / mlb

Source artifact: mlb_shrinkage

Are our probabilities honest?

Reliability diagrams and Murphy decomposition, including the observed information gap.

as of 2026-09-16 / all

Source artifact: murphy_decomposition

The life of a forecast

Observed pre-game information arrival and in-game score checkpoints.

as of 2026-07-25 / all

Source artifact: novel_line_half_life, info_arrival_curve

Home advantage, decomposed

International matches split on the observed neutral-site flag.

as of 2026-07-25T03:52:27.217372+00:00 / soccer

Source artifact: soccer_home_advantage

How accurate were the market sources?

Proportional-devig Brier measurements on shared-game subsets.

as of 2026-07-25T04:01:46.909067+00:00 / all

Source artifact: bookmaker_accuracy

Who bends the shot chart

A transparent NBA on/off rim-deterrence leaderboard with its roster confound.

as of 2026-07-25T04:12:47.517923+00:00 / nba

Source artifact: rim_deterrence

How competitive is each season?

A three-season NBA win-share concentration ledger with scope caveats.

as of date not published / nba

Source artifact: league_parity_index

Greater than the sum of their parts

A five-man residual ledger published with its small-minutes limitation.

as of 2026-07-25T04:25:58.801043+00:00 / nba

Source artifact: lineup_synergy

Is the market calibrated?

A favorite-longshot audit of observed market calibration by sport.

as of 2026-07-25T04:34:37.481061+00:00 / all

Source artifact: market_favorite_longshot

Findings