Every prediction is saved before the match and compared with the result.
In short
Over 36 matches the model got the outcome (1, X or 2) right 61% of the time. Always picking a home win would be right about 59% of the time. Its percentages are 7% less wrong than those taken from league averages alone.
With 36 matches the picture is indicative; it becomes solid at around 100.
How the last round went
20 October – 21 October · outcome right 11 out of 18 (61%) · exact score right 1 time
The surprises
The best predictions
22 correct predictions out of 36. Last 12: 7 correct. Recent form is above the historical average.
Recent form
The latest matches are above the historical average.
Lab · look back
36 predictionsComparison on predictions actually saved before the result. No prediction is rebuilt after the fact.
Model quality
The model is 7% less wrong than the simple method. With 36 matches the figure is still shaky: it needs at least fifty matches, ideally a hundred.
When I say 60%, does it happen?
Each point is a band of predictions (the number next to it says how many matches). The closer the points are to the dashed line, the more reliable the percentages.
If "Stated" and "Actual" are close, the percentages are reliable.
New model or old?
Over 36 matches, statistics only: error 0.41 for the current model, 0.47 for the first version. The current model does better.
Reliability by market
A 0–100 indicator built from calibration and Brier score. It is descriptive, not a guarantee of results. The model's picks use the history only when there are at least 10 observations per category.
Predicted vs actual curve
Each point compares the average stated probability with the actual frequency. The dashed diagonal shows perfect calibration.
Historical calibration active for 5 categories.