How well did our predictions hold up?
We compare the percentages we gave on matches played in the last 8 months with the real results; the report is refreshed every Monday. The number in the table is an error score: the lower it is, the closer our percentages are to reality.
green: better than writing down the league's past average · yellow: only better than a coin toss (equal chance for every result) · red: worse than a coin toss too
What do the terms mean?Error score · League avg. · Coin toss…
- Error score
- In statistics the “Brier score”: the average gap between the percentages we gave and the real result. 0 = perfect, low = good. On the match result, saying 33% for every result scores 0.667.
- League avg.
- The real benchmark: the error score of a simple prediction that writes the league's past result distribution for every match. If our score is below it, judging every match separately is working.
- Coin toss
- An easier bar: equal chance (33%) for every result on the match result, 50%-50% on two-way probabilities. The error scores are 0.667 and 0.500.
- Confidently wrong
- Log-loss. It measures error but punishes being very confident and wrong more heavily; again low = good.
- Green · yellow · red
- Green: better than the league average. Yellow: only better than a coin toss. Red: worse than a coin toss.
2026-W44 — OVERVIEW
Total matches
1122
last 8 months
Error · Match result
0.595
league avg. 0.641 · coin toss 0.667
Error · 3+ goals
0.477
league avg. 0.486 · coin toss 0.500
Error · Both teams score
0.487
league avg. 0.486 · coin toss 0.500
ERROR SCORES LEAGUE BY LEAGUE
PLPremier League
183 matches
- Match result
- 0.614league avg. 0.655
- 3+ goals
- 0.489league avg. 0.492
- Both teams score
- 0.492league avg. 0.495
Confidently wrong: 0.796
PDLa Liga
224 matches
- Match result
- 0.588league avg. 0.633
- 3+ goals
- 0.475league avg. 0.499
- Both teams score
- 0.477league avg. 0.481
Confidently wrong: 0.776
SASerie A
194 matches
- Match result
- 0.574league avg. 0.637
- 3+ goals
- 0.498league avg. 0.500
- Both teams score
- 0.497league avg. 0.500
Confidently wrong: 0.783
BL1Bundesliga
160 matches
- Match result
- 0.588league avg. 0.648
- 3+ goals
- 0.440league avg. 0.447
- Both teams score
- 0.464league avg. 0.462
Confidently wrong: 0.758
FL1Ligue 1
162 matches
- Match result
- 0.607league avg. 0.654
- 3+ goals
- 0.480league avg. 0.485
- Both teams score
- 0.484league avg. 0.487
Confidently wrong: 0.787
CLChampions League
25 matches
- Match result
- 0.581league avg. 0.602
- 3+ goals
- 0.486league avg. 0.461
- Both teams score
- 0.447league avg. 0.435
Confidently wrong: 0.766
TR1Süper Lig
174 matches
- Match result
- 0.607league avg. 0.629
- 3+ goals
- 0.478league avg. 0.489
- Both teams score
- 0.515league avg. 0.499
Confidently wrong: 0.798
| League | Matches | Match result | 3+ goals | Both teams score | Confidently wrong |
|---|---|---|---|---|---|
| PLPremier League | 183 | 0.614league avg. 0.655 | 0.489league avg. 0.492 | 0.492league avg. 0.495 | 0.796 |
| PDLa Liga | 224 | 0.588league avg. 0.633 | 0.475league avg. 0.499 | 0.477league avg. 0.481 | 0.776 |
| SASerie A | 194 | 0.574league avg. 0.637 | 0.498league avg. 0.500 | 0.497league avg. 0.500 | 0.783 |
| BL1Bundesliga | 160 | 0.588league avg. 0.648 | 0.440league avg. 0.447 | 0.464league avg. 0.462 | 0.758 |
| FL1Ligue 1 | 162 | 0.607league avg. 0.654 | 0.480league avg. 0.485 | 0.484league avg. 0.487 | 0.787 |
| CLChampions League | 25 | 0.581league avg. 0.602 | 0.486league avg. 0.461 | 0.447league avg. 0.435 | 0.766 |
| TR1Süper Lig | 174 | 0.607league avg. 0.629 | 0.478league avg. 0.489 | 0.515league avg. 0.499 | 0.798 |
The real benchmark is “league avg.”: a simple prediction that writes the league's past result distribution for every match. Only if we are below this value does judging every match separately add anything. A coin toss (33-33-33% on the match result → 0.667, 50-50% on the other two → 0.500) is an easier bar.
HOW IS IT CALCULATED?
Every Monday the system takes the matches played in the last 8 months and, for each, recalculates which percentages we would have given using data known before that match (nothing from after the match is used, so the result is never “cheated” on). The error score (Brier score) is found like this: for each result, the square of the gap between the percentage we gave and what actually happened (happened = 1, didn't = 0) is summed and averaged over all matches; 0 is perfect. A coin toss scores 0.667 on the match result and 0.500 on two-way probabilities. How we calculate and past results league by league for details.
Report generated: 26 Oct 2026, 14:17