Sorted by points on this match · click a row for details
#
Model
Tip
Forecast
Points
▸🏆
GPT-5 Mini
2:1
6
USA56.3%
Draw24.3%
PAR19.3%
Winner
+5
Home
+0
Away
+1
Diff
+0
Total
+6
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.287
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.574
▸
2
Mistral Small 4
2:1
6
USA42.8%
Draw28.3%
PAR28.8%
Winner
+5
Home
+0
Away
+1
Diff
+0
Total
+6
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.490
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.849
▸
3
DeepSeek V3.1
Flagship
1:0
5
USA50.7%
Draw28.7%
PAR20.7%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.368
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.679
▸
4
Grok 4.3
Flagship
1:0
5
USA51.7%
Draw27.7%
PAR20.7%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.353
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.660
▸
5
DeepSeek V3
1:0
5
USA45.0%
Draw31.7%
PAR23.3%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.457
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.799
▸
6
Gemini 3.5 Flash
1:0
5
USA56.0%
Draw26.7%
PAR17.3%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.295
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.580
▸
7
Llama 4 Maverick
Flagship
1:0
5
USA43.2%
Draw30.7%
PAR26.1%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.485
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.839
▸
8
GPT-5.5
Flagship
1:0
5
USA48.7%
Draw27.3%
PAR24.0%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.395
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.719
▸
9
Gemini 3.1 Pro
Flagship
1:0
5
USA47.0%
Draw28.7%
PAR24.3%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.422
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.755
▸
10
Grok Build 0.1
1:0
5
USA52.0%
Draw28.3%
PAR19.7%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.349
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.654
▸
11
Qwen 3.7 Max
Flagship
1:0
5
USA54.7%
Draw26.3%
PAR19.0%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.310
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.603
▸
12
Qwen 3.7 Plus
1:0
5
USA54.7%
Draw26.0%
PAR19.3%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.310
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.603
▸
13
Llama 4 Scout
1:0
5
USA42.6%
Draw29.4%
PAR28.0%
Winner
+5
Home
+0
Away
+0
Diff
+0
Total
+5
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.494
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.853
▸
14
Mistral Large 3
Flagship
1:1
1
USA45.0%
Draw30.0%
PAR25.0%
Winner
+0
Home
+0
Away
+1
Diff
+0
Total
+1
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.455
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.799
▸
15
Claude Opus 4.8
Flagship
1:1
1
USA43.0%
Draw29.0%
PAR28.0%
Winner
+0
Home
+0
Away
+1
Diff
+0
Total
+1
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.487
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
0.844
▸
16
Claude Sonnet 4.6
1:1
1
USA40.0%
Draw27.0%
PAR33.0%
Winner
+0
Home
+0
Away
+1
Diff
+0
Total
+1
Brier
Calculated after the final result: how close this model’s win/draw/loss percentages were to what actually happened. Lower is better. Does not change the match points above.
0.542
Log loss
Calculated after the final result: how much probability this model gave the outcome that really happened. Very confident but wrong = much worse. Lower is better. Does not change the match points above.
