Upset index
Upset index
Including the years before odds: the biggest upsets in history, the most chaotic seasons and the giant killers by our learning Elo model's pre-match probabilities. Index 100 = as many upsets as expected.
- Matches with a clear favourite
- 614,667
- Upsets
- 126,577
- Upset rate
- 20.6%
- Upset index
- 103
Biggest upsets this week
Last 7 days, all sports
-
Norwegian Football Cup
Rosenborg BK1
HamKam2
Winner's pre-match chance 6.2% · odds 15.00
-
American football
NCAA Division FBS
UCF Knights9
Baylor Bears41
Winner's pre-match chance 6.8% · odds 14.00
-
Puchar Polski · Round of 32
Siarka Tarnobrzeg1
Lech Poznań0
Winner's pre-match chance 6.9% · odds 13.00
Biggest upsets this season
Current season of each competition, by odds
Filters
Model data: 2,438,710 matches with a clear favourite between 26 Apr 1970 and 2 Nov 2026 (learning Elo, pre-match state).
Biggest upsets
By the winner's pre-match chance (Elo model) — lowest first
-
Liga F Moeve
Fútbol Club Barcelona1
Levante UD2
Winner's pre-match chance 1.9%
-
Première Ligue, Féminine
Bordeaux2
OL Lyonnes1
Winner's pre-match chance 2.8%
-
Türkiye Kupası
Beyoğlu Yeni Çarşı FAŞ1
Gaziosmanpaşa SK3
Winner's pre-match chance 2.8%
-
Liga MX, Women, Clausura
Monterrey3
FC Juárez4
Winner's pre-match chance 3.0%
-
Coupe de la Ligue · Quarterfinals
Paris Saint-Germain1
Guingamp2
Winner's pre-match chance 3.1%
-
National League
Telford United1
Lincoln City2
Winner's pre-match chance 3.3%
-
Internacional
1–2
Pelotas
-
Plymouth
1–0
Liverpool FC
-
FC Anry Vladivostok
1–2
Dinamo Barnaul
-
CD Guadalajara (F)
1–2
Mazatlán FC
-
SFK 2000 Sarajevo
2–3
FC Slovan Liberec
-
Çankaya FK
0–3
Hacettepe SK
-
Polideportivo El Ejido 1969
0–1
Sevilla Atlético
-
1461 Trabzon
1–2
Arsinspor
-
Körfez
2–1
Düzcespor
-
Club Atlético de Madrid
0–1
Santa Teresa
-
Grêmio
1–2
EC Cruzeiro
-
Al-Ain
0–1
Al-Fujairah
-
Real Madrid
0–1
Real Betis
-
Fútbol Club Barcelona
2–3
Sporting Huelva
-
PSV Eindhoven
0–2
SC Telstar
-
Vitória
1–4
Itabuna EC
-
Sport Recife
0–1
América-PE
-
Club Sporting Cristal
0–2
Comerciantes Unidos
-
Chelsea
2–4
Bradford City
-
PAOK
0–1
OFPF Ethnikos
-
RFS
2–3
SK Super Nova
-
FC Juárez
0–1
Mazatlán FC
-
Real Madrid
0–1
SD Eibar
-
Athletico
0–1
Oeste Brasil
-
Olympiacos FC
0–1
AOK Kerkyra
-
Oxford United
0–1
Stockport County
-
Atlético de San Luis
1–0
Monterrey
-
União Recreativa dos Trabalhadores
1–0
Atlético-MG
-
FC Bayern München
1–2
FC Augsburg
-
APO Ellas Syrou
0–1
Olympiacos B
-
Goiás
0–2
Brasília
-
Madrid CFF
2–1
Fútbol Club Barcelona
-
Avaí
0–2
Concórdia AC
-
Manchester United
2–3
Blackburn
Upset index by league
100 = as many upsets as the probabilities predict; above is more chaotic, below more predictable
Giant killers
Win more than expected as the underdog
| Team | Underdog wins | Expected | Index |
|---|---|---|---|
| 18/31 | 7.2 | 205 | |
| 10/19 | 4.3 | 172 | |
| 37/99 | 22.7 | 171 | |
| 9/13 | 3.7 | 171 | |
| 15/29 | 8.0 | 168 | |
| 26/62 | 15.6 | 168 | |
| 19/44 | 10.9 | 166 | |
| 10/14 | 4.7 | 166 | |
| 34/95 | 21.7 | 164 | |
| 9/17 | 4.2 | 163 |
Upset victims
Lose more than expected as the favourite
| Team | Losses as favourite | Expected | Index |
|---|---|---|---|
| 12/18 | 4.9 | 184 | |
| 23/49 | 11.9 | 184 | |
| 10/15 | 3.7 | 184 | |
| 14/25 | 6.3 | 182 | |
| 26/58 | 14.2 | 180 | |
| 16/34 | 7.9 | 178 | |
| 38/93 | 22.4 | 177 | |
| 11/17 | 4.7 | 177 | |
| 10/20 | 4.1 | 176 | |
| 14/28 | 6.8 | 175 |
How is it calculated?
- Probability: From pre-match 1X2 (or two-way) odds, with the bookmaker margin removed proportionally: p = (1/odds) / Σ(1/odds). The same formula as the stats hub's “upsets by odds” block.
- Clear favourite and upset: Only matches with at least a 10-point gap between the home and away win probabilities count; if the favourite's opponent wins, it is an upset. Draws are not upsets. Winner's chance of 25% or less = big upset.
- Ranking: The lower the winner's pre-match chance, the bigger the upset (surprise score = 100 × (1 − chance)).
- Index: 100 × (actual upsets + 5) / (expected upsets + 5); expected = sum of the underdogs' win probabilities. The +5 pulls small samples towards 100. Teams have two directions: giant killing (winning as the underdog) and fragility (losing as the favourite).
- Calibration: Expected values are scaled by the source's overall hit factor (0.86 in this view), so a systematic bias of the source does not shift every league the same way.
- Historical (model): For past matches without odds, the probability comes from our learning Elo model replaying the full history chronologically and taking its state before the match (both teams need at least 15 matches; friendlies excluded). Never mixed with odds.