Our model's five strongest calls of the 22–24 August weekend
Published
Every week our model prices hundreds of outcomes, and most of them live in the messy middle — 40-something percents that honest sites report and dishonest sites round up to “banker”. This weekend, exactly five fixtures cleared a 65% home-win probability in our 16 August model run. Here they are, strongest first, with the current numbers always live on each match page.
The five
- Sporting CP vs Alverca — the strongest call on our board this weekend at 77.6% for the home win, with Over 2.5 goals also favoured at 67%.
- Inter vs Monza — 72.6% for Inter at San Siro on Serie A’s opening weekend.
- PSV vs Groningen — 69.4% for PSV, but the goals market is the louder call here: Over 2.5 at 74%, the highest of these five.
- Palmeiras vs Vasco da Gama — 68.4% for Palmeiras, mid-season Brasileirão with plenty of form data behind it.
- FC Porto vs Arouca — 67.0% for Porto at the Dragão.
The honest math
Here’s the part most tip lists leave out. Multiply those five probabilities together and the chance that all five favourites win is about 18%. Read that again: even our five most confident calls of the entire weekend, combined, are more than four times as likely to produce at least one upset as a clean sweep. That’s not a flaw in the model — it’s what 67–78% actually means, and it’s why we publish probabilities instead of promises.
A 72% call that loses wasn’t “wrong”; it was a 28% event happening, which it should do roughly two times in seven. Whether our percentages are calibrated — whether our 70% calls really win about 70% of the time — is a question you can check yourself on the track record page, where every one of these five will be graded after full time, hits and misses alike.
And the other extreme
The same model run also produced the weekend’s least decided fixture: Newcastle vs Liverpool, where one tenth of a percentage point separates the two sides. We wrote up why the opener at St James’ Park is a genuine coin flip — worth reading next to this list, because a model you can trust on the near-locks is the same model that admits when it has no idea.
