Transparent production metrics
How the CSP model works
The headline contains only the production match-winner model and one frozen prediction per match.
Settled
1233
Accuracy
58.5%
Baseline
57.9%
Brier score
0.238
Log loss
0.669
Quality by confidence
| Stated probability | Settled | Accuracy | Brier |
|---|---|---|---|
| 0.5-0.6 | 805 | 54.7% | 0.246 |
| 0.6-0.7 | 399 | 65.9% | 0.222 |
| 0.7-0.8 | 29 | 62.1% | 0.242 |
Calibration
Predicted probability compared with the observed outcome rate.
0.2-0.3 · n=1
26.2% / 0.0%
0.3-0.4 · n=42
37.4% / 47.6%
0.4-0.5 · n=404
45.1% / 49.5%
0.5-0.6 · n=401
55.3% / 58.9%
0.6-0.7 · n=357
65.6% / 67.5%
0.7-0.8 · n=28
71.5% / 60.7%
Separate from the headline
Experimental models
Map totals and round totals are still being validated and never affect the production winner accuracy shown above.
By market
| Market | Settled | Accuracy | Won / Lost |
|---|---|---|---|
| Map round totals | 1233 | 58.5% | 714 / 519 |
By line
Map round totals
| Line | Settled | Accuracy |
|---|---|---|
| 19.5 | 411 | 66.9% |
| 20.5 | 411 | 57.2% |
| 21.5 | 411 | 51.3% |
Immutable settled history
Prediction archive
No settled predictions for these filters yet.