HOME /Track record

Public results · auto-settled · no betting advice

Season settled951
Best win rate52.7%
Avg Brier0.2705
Calibration n930

Season trust signals

MLBelo-v1n=905
Brier0.2493Near coin flip
Current streakW4since Jul 20
NBAelo-v1n=25
Brier0.2917Worse than a coin flip (honest signal)
Current streakL1since Jul 3

Week highlights (last 7 days)

Home vs away breakdown (moneyline)

Splits our home-side picks from our away-side picks. If one side wins markedly more than the other, the model is mis-pricing home-court advantage.

MLBAway win
Win rate57.8%
Hits / attempts26 / 45
MLBHome win
Win rate52.4%
Hits / attempts451 / 860
NBAHome win
Win rate40.0%
Hits / attempts10 / 25

Last 7 days

Recent trend

Moneyline · win rate

LeagueModelWin rateHits / attemptsSettledAvg confidence
MLBelo-v157.8%26 / 454553.0%

Total · MAE error

No total settlements yet

Last 30 days

Monthly trend

Moneyline · win rate

LeagueModelWin rateHits / attemptsSettledAvg confidence
MLBelo-v152.3%224 / 42843953.4%
NBAelo-v1Sample too small (n=19)3 / 91959.8%

Total · MAE error

No total settlements yet

Season

All-time

Moneyline · win rate

LeagueModelWin rateHits / attemptsSettledAvg confidence
MLBelo-v152.7%477 / 90591653.5%
NBAelo-v140.0%10 / 253561.3%

Total · MAE error

No total settlements yet

World Cup 1X2 (90-minute win-draw-loss)

Brier is a three-class normalized score (0–1, lower is better); not yet benchmarked against a favourite baseline or market odds.

1X2 · win rate

LeagueModelWin rateHits / attemptsBrierSettled
FIFA World Cupwc-elo-v162.0%57 / 920.256792

Secondary markets (season)

Spread Brier uses the model's cover probability (0.25 ≈ coin flip, lower is better). Same auto-settled data as above, published unfiltered.

Spread · win rate + Brier

No spread settlements yet

First-half winner · win rate

No first-half winner settlements yet

First-half + team totals · MAE error

No first-half / team total settlements yet

Perfect calibration

Confidence calibration

When we say '70% to win,' how often do those teams actually win? Points hugging the dashed line mean the model's confidence is honest. Blue = beats prediction (under-confident), red = worse than prediction (over-confident), green = within ±5%.

Perfect calibration50%60%70%80%90%100%0%25%50%75%100%n=915n=15Predicted confidenceActual hit rate

Moneyline model · 930 settled games (voids excluded)