What we publish
For each match: the chance of a home win, a draw and an away win, with a 90% range around each number. This weekend that is fifteen MLS matches and six Liga MX matches, from 31 October to 1 November 2026, all on the forecasts page.
The numbers come from team ratings built only from this season's results (480 MLS regular-season matches, and 281 Liga MX league matches this year), turned into probabilities by a simple statistical model fitted to the same season. No club data and no video go into these public forecasts.
For FC Cincinnati vs Philadelphia Union we also recorded the range from our pre-game report, made with two separate methods: Cincinnati 35-36%, draw 21-28%, Philadelphia 36-44%, with expected goals of 1.84-2.22 for Cincinnati against 2.05-2.19 for Philadelphia.
How we keep score
Every forecast is saved with the time it was made, before kick-off. After the final whistle, a script reads the final score from a public results feed and scores the forecast. Nobody types results in by hand, and nothing is deleted afterwards: the misses stay next to the hits on the track record.
We use two standard scores. The Brier score adds up the squared gap between each probability and what happened. The ranked probability score (RPS) does the same but treats home win, draw and away win as an ordered scale, so a near miss costs less than a wide one. For both, lower is better.
Each forecast is also scored against a benchmark: the season's base rates, meaning how often home wins, draws and away wins happen. A model is only useful if it beats that.
Why one weekend proves very little
Twenty-one matches is a tiny sample. A 40% chance should fail more often than it comes true. A good model will have bad weekends and a poor one will have lucky ones. The test that matters is a full season: hundreds of forecasts, scored the same way, against the same benchmark.
Our own test in MLS, in date order on 292 earlier matches, gave the ratings a small edge over base rates: an RPS of 0.217 against 0.222. The gap points the right way but is not yet statistically clear. We will keep publishing the running average, whether it flatters us or not.
Why this matters to a club
The same rule applies to every report we make for a partner club: forecasts recorded before the match and checked after it. Public ratings are only the starting point. A club's own statistics, GPS and video are where a real advantage comes from, and that work is private to the club.