Skip to content
Browse › Soccer Calibration Pack
Analytics module · as of

Soccer Calibration Pack

soccer_intl corpus: n=9003 rows -- far smaller than mlb/nba, wider CIs, single-fold reads not durable. Murphy: model Brier 0.2279 vs market 0.1427 (gap=+0.0852); market leads. Minute-bucket ECE (weighted): 0.3580.
View full size ↗
confirmednull (a finding)not testabledescriptivepending
Soccer Calibration Pack
Chart: Soccer Calibration Pack -- soccer_intl corpus: n=9003 rows -- far smaller than mlb/nba, wider CIs, single-fold reads not durable. Murphy: model Brier 0.2279 vs market 0.1427 (gap=+0.0852); market leads. Minute-bucket ECE (weighted): 0.3580.
scripts/platformkit/analytics_showcase/out/soccer_calibration_pack.json

What it means

When model and market diverge by 0.10 or more (n=4406), the model is the closer forecast only 21.52% of the time. Late-game calibration also drifts, with expected calibration error climbing to 0.4978 in the 75-91 minute bucket. This is a published null, not a win.

Caveats & confounds

The soccer_intl corpus is only 9,003 rows, far smaller than MLB or NBA, so confidence intervals are wide and single-fold reads are not durable.

Ask Scout about this
How good is the soccer model's calibration?Are single-fold results trustworthy?Ask anything →