Upset Tracker
Matches the model-underdog won — shown with the win probability we published before kickoff — alongside the upcoming fixtures with the most upset potential. A forecast isn’t wrong when the underdog wins; this is about understanding where the game is most open.
Upsets to Watch
Upcoming matches with a clear favourite but a live underdog
No upcoming high-uncertainty matches right now.
Upsets So Far
Completed matches the model-underdog won, with the win probability we published
Group AJun 25
South Africa beat South Korea0 - 1We gave South Africa 13.9% to winSurprisal 86.1
Group EJun 14
Ivory Coast beat Ecuador1 - 0We gave Ivory Coast 14.7% to winSurprisal 85.3
Group LJun 17
Ghana beat Panama1 - 0We gave Ghana 17.9% to winSurprisal 82.1
KnockoutJul 5
Norway beat Brazil1 - 2We gave Norway 24.2% to winSurprisal 75.8
Group DJun 26
Turkey beat United States2 - 3We gave Turkey 25.3% to winSurprisal 74.7
Group DJun 13
United States beat Paraguay4 - 1We gave United States 26.2% to winSurprisal 73.8
KnockoutJun 29
Paraguay beat Germany1 - 1 (3-4 pens)We gave Paraguay 26.2% to winSurprisal 73.8
KnockoutJul 3
Egypt beat Australia1 - 1 (2-4 pens)We gave Egypt 26.5% to winSurprisal 73.5
KnockoutJul 7
Switzerland beat Colombia0 - 0 (4-3 pens)We gave Switzerland 26.7% to winSurprisal 73.3
Group DJun 14
Australia beat Turkey2 - 0We gave Australia 27.9% to winSurprisal 72.1
Group EJun 25
Ecuador beat Germany1 - 2We gave Ecuador 28.4% to winSurprisal 71.6
Group KJun 27
Congo DR beat Uzbekistan1 - 3We gave Congo DR 31.6% to winSurprisal 68.4
Group DJun 20
Paraguay beat Turkey1 - 0We gave Paraguay 31.7% to winSurprisal 68.3
KnockoutJun 30
Morocco beat Netherlands1 - 1 (2-3 pens)We gave Morocco 33.7% to winSurprisal 66.3
Group DJun 19
United States beat Australia2 - 0We gave United States 35.4% to winSurprisal 64.6
How our forecasts calibrate
Across 103 scored matches: how often outcomes happened, grouped by the probability we gave them
We said ~13%happened 11% (n=70)
We said ~28%happened 22% (n=152)
We said ~49%happened 66% (n=50)
We said ~69%happened 78% (n=32)
We said ~84%happened 60% (n=5)
Published probabilities are frozen at kickoff — the number we showed, never adjusted after the result. This page is for understanding the model, not for wagering.