검증하기
확률은 실현되고 있는가?
모델이 어떤 팀에 70% 확률을 부여하면, 그 팀은 열 번 중 약 일곱 번 이겨야 합니다. 이 페이지는 그렇게 되고 있는지 확인합니다. 2026 월드컵의 모든 경기는 끝나는 즉시 채점되며, 아래 수치는 한 가지 질문에 답합니다: 확률이 자신감만 있는 것이 아니라 정직한가? (이것의 기술적 명칭이 보정입니다.)
Track record
Proven on past tournaments
The short version: when the model says 70%, it happens about 70%. Across these 987 matches its stated chances landed within ~5.6 points of reality, and on average it rated the actual result about 35% more likely than a blind 1-in-3 guess would.
The 2026 tracker below stays empty until the first match kicks off. So to show the model has been tested, not just described, we ran it against tournaments whose results you already know. For each one, the model was rebuilt exactly as it stood the day before kickoff, then graded on every match — it never sees the result it is being marked on. That is what “graded out of sample” means: no peeking, no hindsight.
Each tournament is scored by the production model reconstructed as it stood the day before the tournament's first match: Dixon-Coles and Hierarchical Poisson refit on matches strictly beforehand, Elo rolled forward to each match, and the tournament-tier calibration layer refit on the 24 months of matches before the cutoff. No data from the tournament, or any later match, touches any layer of the fit.
Graded across 987 matches at 24 major tournaments (2014–2024)
- 0.572
- 1.000
- 5.6pp
How close the forecasts landed to reality. Lower is better; blind 1-in-3 guessing scores ≈ 0.667.
Like Brier, but overconfidence is punished harder. Lower is better; blind guessing scores ≈ 1.099.
Does “70%” really mean 70%? The average gap between the two. Lower is better.
Tournament by tournament
One row per tournament: the model rebuilt as it stood the day before it began, then graded on every match through the final (Brier — lower is better, blind 1-in-3 guessing is 0.667). A few thin, early editions sit above that line; the honest measure is all of them pooled, in the box above.
| Tournament | Host | Matches | Brier |
|---|---|---|---|
| Copa América 2024 | United States | 32 | 0.522 |
| Euro 2024 | Germany | 51 | 0.613 |
| AFCON 2024 | Ivory Coast | 52 | 0.651 |
| Asian Cup 2024 | Qatar | 51 | 0.515 |
| Gold Cup 2023 | United States | 31 | 0.566 |
| World Cup 2022 | Qatar | 64 | 0.611 |
| AFCON 2022 | Cameroon | 52 | 0.686 |
| Gold Cup 2021 | United States | 31 | 0.341 |
| Copa América 2021 | Brazil | 28 | 0.481 |
| Euro 2021 | England | 51 | 0.554 |
| AFCON 2019 | Egypt | 52 | 0.546 |
| Gold Cup 2019 | United States | 31 | 0.405 |
| Copa América 2019 | Brazil | 26 | 0.542 |
| Asian Cup 2019 | United Arab Emirates | 51 | 0.496 |
| World Cup 2018 | Russia | 64 | 0.569 |
| Gold Cup 2017 | United States | 25 | 0.456 |
| AFCON 2017 | Gabon | 32 | 0.642 |
| Euro 2016 | France | 51 | 0.668 |
| Copa América 2016 | United States | 32 | 0.502 |
| Gold Cup 2015 | United States | 26 | 0.755 |
| Copa América 2015 | Chile | 26 | 0.686 |
| AFCON 2015 | Equatorial Guinea | 32 | 0.795 |
| Asian Cup 2015 | Australia | 32 | 0.434 |
| World Cup 2014 | Brazil | 64 | 0.565 |
Reliability diagram
Read it like this: each dot is a group of similar forecasts. left-to-right is what the model predicted, bottom-to-top is how often it actually happened. When the model says 70% and that happens about 70% of the time, the dot sits on the dashed line: perfect calibration. The closer the dots hug the line, the more honest the probabilities; bigger dots mean more matches in that group.
Brier by favourite confidence
Matches grouped by how confident the model's favourite was (its biggest of the home / draw / away probabilities) — so you can see whether it is as reliable on toss-ups as on heavy favourites.
| Favourite confidence | Matches | Brier |
|---|---|---|
| P_fav < 40% | 81 | 0.649 |
| P_fav 40-60% | 476 | 0.633 |
| P_fav 60-80% | 318 | 0.512 |
| P_fav >= 80% | 112 | 0.428 |
- Out-of-sample: the calibration layer is refit per tournament on pre-tournament data, so these numbers do not reuse the live shipped calibrator (which has seen these results).
- The uniform 1/3 forecast scores a Brier of 0.667; lower is better. Major-tournament football is high-variance, so a strong model still sits well above a league-season Brier.
- Calibrated and uncalibrated metrics are reported on the same fixtures so the calibration layer's effect is visible.
Built 2026-05-30 · model 1.0.0 · calibration layer refit on the 24 months before each tournament.
2026 월드컵 — 실시간 트래커
Across 96 graded forecasts so far, the model's overall Brier score is 0.497 (lower is better; blind 1-in-3 guessing scores ≈ 0.667). The reliability diagram below plots what the model predicted against how often it actually came true. The closer to the diagonal, the more honest the probabilities.
Overall, across 96 scored matches
- 0.497
- 0.837
- 0.167
How close the forecasts landed to reality. Lower is better; blind 1-in-3 guessing scores ≈ 0.667.
Like Brier, but overconfidence is punished harder. Lower is better; blind guessing scores ≈ 1.099.
Does “70%” really mean 70%? The average gap between the two. Lower is better.
Rolling Brier, 10-match window
The model's accuracy over its most recent 10 matches, re-figured after each game. Lower is better; the dashed line is what blind 1-in-3 guessing would score, so anything below it is real skill.
Per-matchday breakdown
| Date | Matches | Brier | Log loss |
|---|---|---|---|
| 2026-06-11 | 2 | 0.404 | 0.723 |
| 2026-06-12 | 2 | 0.817 | 1.280 |
| 2026-06-13 | 4 | 0.771 | 1.178 |
| 2026-06-14 | 4 | 0.595 | 1.008 |
| 2026-06-15 | 4 | 1.148 | 1.632 |
| 2026-06-16 | 4 | 0.197 | 0.438 |
| 2026-06-17 | 4 | 0.640 | 1.005 |
| 2026-06-18 | 4 | 0.426 | 0.739 |
| 2026-06-19 | 4 | 0.463 | 0.787 |
| 2026-06-20 | 4 | 0.531 | 0.854 |
| 2026-06-21 | 4 | 0.603 | 0.928 |
| 2026-06-22 | 4 | 0.266 | 0.529 |
| 2026-06-23 | 4 | 0.433 | 0.706 |
| 2026-06-24 | 6 | 0.410 | 0.750 |
| 2026-06-25 | 6 | 0.506 | 0.868 |
| 2026-06-26 | 6 | 0.388 | 0.704 |
| 2026-06-27 | 6 | 0.434 | 0.763 |
| 2026-06-29 | 2 | 0.612 | 0.996 |
| 2026-06-30 | 3 | 0.500 | 0.866 |
| 2026-07-01 | 3 | 0.478 | 0.830 |
| 2026-07-02 | 3 | 0.327 | 0.622 |
| 2026-07-03 | 3 | 0.460 | 0.796 |
| 2026-07-04 | 1 | 0.225 | 0.470 |
| 2026-07-06 | 2 | 0.443 | 0.784 |
| 2026-07-07 | 2 | 0.541 | 0.909 |
| 2026-07-09 | 1 | 0.573 | 0.960 |
| 2026-07-10 | 1 | 0.280 | 0.564 |
| 2026-07-11 | 2 | 0.250 | 0.517 |
| 2026-07-15 | 1 | 0.537 | 0.911 |
Updated 2026-07-16.
Reliability diagram
Read it like this: each dot is a group of similar forecasts. left-to-right is what the model predicted, bottom-to-top is how often it actually happened. When the model says 70% and that happens about 70% of the time, the dot sits on the dashed line: perfect calibration. The closer the dots hug the line, the more honest the probabilities; bigger dots mean more matches in that group.
Brier by competition
| Segment | Matches | Brier | Δ vs overall |
|---|---|---|---|
| World Cup 2026 | 96 | 0.497 | +0.000 |
Brier by tournament stage
| Segment | Matches | Brier | Δ vs overall |
|---|---|---|---|
| Group stage | 72 | 0.516 | +0.019 |
| Round of 32 | 15 | 0.449 | -0.048 |
| Round of 16 | 4 | 0.492 | -0.005 |
| Quarter-final | 4 | 0.338 | -0.159 |
| Semi-final | 1 | 0.537 | +0.039 |
| Third-place play-off | — | — | — |
| Final | — | — | — |
Brier by favourite confidence
| Segment | Matches | Brier | Δ vs overall |
|---|---|---|---|
| P_fav < 40% | 11 | 0.669 | +0.172 |
| P_fav 40-60% | 47 | 0.540 | +0.043 |
| P_fav 60-80% | 32 | 0.373 | -0.125 |
| P_fav >= 80% | 6 | 0.513 | +0.016 |
세 가지 수치의 의미
날씨 예보관을 생각해 보세요. 누구나 "비 올 확률 70%"라고 말할 수 있습니다. 실력 있는 예보관은 그렇게 말했을 때 실제로 약 70%의 확률로 비가 옵니다. 이 세 가지 수치가 모델을 같은 방식으로 점검합니다.
- Brier score: 확률이 현실에 가까웠는가? 매 경기마다 예측이 실제 결과와 얼마나 떨어져 있었는지 측정한 뒤 평균을 냅니다. 완벽한 예측은 0점, 맹목적으로 매번 1/3을 찍으면 약 0.667점입니다. 낮을수록 좋습니다.
- Log loss: 같은 개념이지만, 과신에 엄격한 벌점을 줍니다. 거의 확실하다고 한 뒤 틀리면 이 수치가 급등합니다. 모델을 겸손하게 유지하는 지표입니다. 맹목적 추측은 약 1.099점입니다. 낮을수록 좋습니다.
- ECE: "70%"는 정말 70%를 의미하는가? "약 70%" 예측을 모두 모아서 실제로 얼마나 자주 일어났는지 확인합니다. 모든 신뢰 수준에 걸친 평균 격차가 ECE입니다. 몇 퍼센트포인트 이내라면 명시된 확률을 그대로 신뢰할 수 있습니다. 낮을수록 좋습니다.
처음 두 개는 정확하면서도 대담한 것에 보상을 주고, 마지막은 정직성 점검입니다. 모델이 인상적으로 보이면서도 자신감을 과대 표현할 수 있으므로, 세 가지를 함께 측정해야 그것을 잡아낼 수 있습니다.
이 수치 뒤의 구성 모델과 표본 외 테스트를 알고 싶으시면 방법론 페이지에서 확인하세요.