DiamondIQ · Transparency

Every prediction we've ever made, graded

Our win-probability model logs a prediction for every game — pre-game, at the closing line, and at each half-inning while it's live — and grades itself once the game is final. No windows are cherry-picked; losses are shown with the same prominence as wins. Updated daily.

Games graded
96
Brier score
0.2506
lower is better
vs coin flip (.2500)
+0.0006
Brier improvement
vs market close
Behind
64 games w/ a line

Calibration — Predicted Probability vs Actual Win Rate

00252550507575100100Predicted home win %

A perfectly calibrated model sits on the diagonal: when we say 70%, the home team should win about 70% of the time. Dots are sized by how many games fall in each 10-point bucket.

PredictedGamesAvg predActual win
30–40%140%100%
40–50%3148%35%
50–60%5755%61%
60–70%663%33%
70–80%171%0%

Brier Score vs Baselines — Lower Is Better

DiamondIQ model
0.2506
Market closing line · over 64 games with a line
0.2482
Always 50% · a pure coin flip
0.2500
Apples-to-apples on the 64 games that had a market line: DiamondIQ 0.2520 vs market 0.2482. Log loss — DiamondIQ 0.6949, market 0.6915, coin flip 0.6931.

Head to Head vs the Market Closing Line

Beat the market's Brier
25–39
of 64 games
When we disagreed on the pick
6–6
12 divergent games
Market games
64
Live predictions logged
988
"Divergent" = games where our model favored the opposite side of the market's favorite. This is context, not a betting claim — see the methodology note below on why the comparison is not yet apples-to-apples.
Historical validation · backtest

Historical Validation (Backtest) — Retrospective Research

This is retrospective research, not the live record. The numbers below come from a 9,719-game historical replay (2022–2025) used to validate model v2 before it was deployed. They are separate from — and not part of — the graded live predictions shown above.

Methodology. An as-of-date replay: each game is re-predicted using only data that existed before first pitch (records, probable starters, PitchIQ as of that day). The v2 pitcher adjustment and calibration were fit on 2022–2024 (7,289 games); the 2025 season (2,430 games) was held out from all tuning, so the holdout numbers are the honest out-of-sample test.

Holdout — 2025 (2,430 games), v1 vs v2
Metricv1 (records-only)v2 (deployed)
Brier score · lower is better0.24980.2452
Calibration error (ECE) · lower is better0.05920.0226
Log loss · lower is better0.69360.6832
Accuracy · picks vs actual55.5%55.3%
Coin-flip Brier = 0.2500. On the untouched 2025 holdout, v2 improved both Brier (−0.0047) and calibration (ECE nearly halved), so it was cleared to replace v1. This backtest cannot use live-only inputs (Statcast arsenal/velocity), and it validates the pre-game prior only — the in-game win-expectancy is unchanged. Full gating methodology lives in the repository report REPORT_V2.md (not web-published).

Methodology & Honesty

The model (v2). DiamondIQ estimates home win probability from team records and home-field advantage (a log5 prior), the starting-pitcher quality gap (PitchIQ), and a calibration shrink — with live game state once a game is underway. The pitcher adjustment and calibration were fit on 2022–2024 and validated on the untouched 2025 season (v2 beat v1 on both Brier and calibration). It does not yet model bullpens, lineups, or weather.

Every prediction is version-tagged. Rows are labeled with the model version that produced them (v1 → v2), so this track record shows the transition honestly rather than silently swapping models under a single number.

Why the market comparison is still generous to the market. The closing line also prices in bullpens, lineups, and late scratches that v2 does not. We show the comparison for accountability, not to claim an edge.

How grading works. Every prediction is scored with the Brier score (p − outcome)² and log loss once the game is final. Calibration buckets predictions by predicted probability and compares each bucket to the actual home win rate. Nothing is re-stated or re-scored after the fact.

No cherry-picking. The numbers above are the cumulative record over every graded game — no date windows, no dropped losses. The full log is downloadable as CSV.

⬇ Download the full prediction log (CSV)
Every logged prediction with its market line, game state, and grade. Updated daily · covering Jul 6, 2026–Jul 16, 2026.
DiamondIQ · Newsletter

Get the numbers in your inbox

Double opt-in. Unsubscribe any time. We never share your email.