Understanding Football Probability
Why the most likely outcome is never the only outcome.
Why the most likely outcome is never the only outcome.
Three outcomes, one distribution
A full-time football forecast normally partitions results into home win, draw and away win. These events are mutually exclusive and collectively exhaustive, so their probabilities must sum to 100%. The exact settlement scope, including how abandoned games are treated, must be declared before evaluation.
Most likely does not mean certain
In an illustrative 45% / 30% / 25% forecast, the home win is the largest individual outcome. Yet there is still a 55% chance of another result. Describing the home side as the most likely winner is valid; describing a win as inevitable is not.
Think in repeated forecasts
One result cannot demonstrate calibration. Across many comparable forecasts near 65%, a calibrated model would see the event happen roughly 65% of the time. The observed proportion will vary in small samples, and comparisons should show how many predictions each group contains.
Use the full distribution
Read all three outcomes, the timestamp and the input assumptions. Model disagreement can identify sensitivity to assumptions, but it is not a confidence interval. The illustrative confidence index shown in the NinetyQuant demo is neither a win probability nor an empirically validated reliability measure.
Sources & further reading
Prepared by NinetyQuant. Numerical examples are illustrative. References explain general concepts and do not validate NQ’s demo forecasts.