Back to board
2026-W38 · Sep 14 – Sep 20, 2026
Weekly accuracy report
Every one of the 46 games that finished this week was re-graded, scoring the 1773 probabilities that had been published against actual outcomes. Lower Brier is better; 0.25 is coin-flip level.
Brier (model)
0.2033
raw 0.2059
Calibration gain
−0.0026
lower is better
Hit rate
68.9%
base 35.3%
Calibration error
3.1%
raw 5.0%
By league
| League | Games | Scored | Brier |
|---|---|---|---|
| MLB | 18 | 1508 | 0.2001/ 0.2038 |
| J리그 | 8 | 104 | 0.2075/ 0.2017 |
| EPL | 5 | 65 | 0.2616/ 0.2599 |
| MLS | 4 | 52 | 0.1948/ 0.1953 |
| CS2 | 11 | 44 | 0.2290/ 0.2223 |
What changed this week
- • Graded 1773 markets from yesterday. Shipped probabilities scored Brier 0.2033 (raw 0.2059; a coin flip is 0.25).
- • MLB: learned calibration cut Brier by 0.0037 versus raw output (1508 samples).
- • batter_rbi_run markets ran 3.1%p low versus outcomes (400 samples, 35% actual). The refit shifts them up.
- • batter_bb markets ran 3.3%p high versus outcomes (200 samples, 25% actual). The refit shifts them down.
- • J리그: calibration trailed raw output by 0.0058; coefficients stay conservative until more samples land.
- • team_totals markets ran 9.1%p high versus outcomes (47 samples, 38% actual). The refit shifts them down.
- • epl, jleague crossed the sample threshold and joined the learning set automatically.
- • CS2: calibration trailed raw output by 0.0067; coefficients stay conservative until more samples land.