Bars: home · draw · away. Heatmap rows = home goals, columns = away goals; brighter = more likely, outlined cell = most-likely score (), green ring = actual result ().
Title odds · who wins the World Cup?
▾Each team's chance to lift the trophy, from 20,000 simulations of the whole tournament — every remaining group match sampled from the model, then the verified knockout bracket played out (ties resolved by extra-time/penalty coin-flip).
1 Argentina
16.0%
F 26%R16 88%2 Brazil
13.5%
F 21%R16 71%3 Spain
10.9%
F 20%R16 73%4 England
9.1%
F 16%R16 74%5 France
7.2%
F 15%R16 73%6 Colombia
6.5%
F 13%R16 81%7 Belgium
5.9%
F 12%R16 67%8 Portugal
5.4%
F 11%R16 65%9 Netherlands
4.4%
F 10%R16 56%10 Germany
4.0%
F 10%R16 81%11 Morocco
2.3%
F 6%R16 44%12 Switzerland
2.3%
F 6%R16 73%13 Ecuador
1.4%
F 3%R16 53%14 Japan
1.3%
F 3%R16 29%15 Norway
1.1%
F 3%R16 59%16 Australia
1.0%
F 3%R16 57%17 Croatia
1.0%
F 3%R16 35%18 Senegal
0.9%
F 3%R16 51%19 Canada
0.9%
F 3%R16 67%20 Algeria
0.8%
F 2%R16 33%21 United States
0.8%
F 2%R16 49%22 Mexico
0.7%
F 2%R16 47%23 Austria
0.7%
F 2%R16 27%24 Egypt
0.4%
F 1%R16 43%25 Sweden
0.4%
F 1%R16 26%26 Ivory Coast
0.4%
F 1%R16 41%27 Paraguay
0.3%
F 1%R16 27%28 Ghana
0.1%
F 0%R16 19%29 South Africa
0.1%
F 0%R16 33%30 DR Congo
0.1%
F 0%R16 27%31 Bosnia and Herzegovina
0.1%
F 0%R16 19%32 Cape Verde
0.0%
F 0%R16 12%33 Curaçao
0.0%
F 0%R16 0%34 Czech Republic
0.0%
F 0%R16 0%35 Haiti
0.0%
F 0%R16 0%36 Iran
0.0%
F 0%R16 0%37 Iraq
0.0%
F 0%R16 0%38 Jordan
0.0%
F 0%R16 0%39 New Zealand
0.0%
F 0%R16 0%40 Panama
0.0%
F 0%R16 0%41 Qatar
0.0%
F 0%R16 0%42 Saudi Arabia
0.0%
F 0%R16 0%43 Scotland
0.0%
F 0%R16 0%44 South Korea
0.0%
F 0%R16 0%45 Tunisia
0.0%
F 0%R16 0%46 Turkey
0.0%
F 0%R16 0%47 Uruguay
0.0%
F 0%R16 0%48 Uzbekistan
0.0%
F 0%R16 0%What the model learned · team strength
▾Latent attack and defense strengths (posterior means, ranked by overall quality). These are the parameters the model fits from goals — the reason behind every projected scoreline. Bars are relative to the WC2026 field.
1 Brazil
2 Argentina
3 Spain
4 England
5 Portugal
6 France
7 Belgium
8 Colombia
9 Netherlands
10 Uruguay
11 Germany
12 Morocco
13 Switzerland
14 Japan
15 Croatia
16 Ecuador
17 Norway
18 Iran
19 Australia
20 Austria
21 Mexico
22 Algeria
23 Senegal
24 United States
25 Canada
26 Sweden
27 South Korea
28 Ivory Coast
29 Turkey
30 Egypt
31 Czech Republic
32 Paraguay
33 Tunisia
34 Scotland
35 Uzbekistan
36 DR Congo
37 New Zealand
38 Ghana
39 Saudi Arabia
40 South Africa
41 Iraq
42 Qatar
43 Panama
44 Bosnia and Herzegovina
45 Jordan
46 Cape Verde
47 Haiti
48 Curaçao
No matches for that team.
Group A
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Mexico9 pts
100%
100% South Africa4 pts
100%
0% Czech Republic1 pts
0%
0% South Korea3 pts
0%
0% Mexico v South Africa
2–0final · predicted 1–0 · ✓ called it
Proj. 1.75–0.62balancedHost (home)
O1.5 68%O2.5 42%O3.5 22%BTTS 38%
South Korea v Czech Republic
2–1final · predicted 1–1 · ✓ called it
Proj. 1.26–1.13balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 48%
Czech Republic v South Africa
1–1final · predicted 1–0 · ✗ missed
Proj. 1.30–0.81balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 40%
Mexico v South Korea
1–0final · predicted 1–0 · ✓ called it
Proj. 1.52–0.96balancedHost (home)
O1.5 71%O2.5 45%O3.5 24%BTTS 48%
Czech Republic v Mexico
0–3final · predicted 1–1 · ✓ called it
Proj. 1.29–1.37balancedHost (home)
O1.5 74%O2.5 49%O3.5 28%BTTS 54%
South Africa v South Korea
1–0final · predicted 0–1 · ✗ missed
Proj. 0.70–1.25balancedNeutral
O1.5 58%O2.5 31%O3.5 13%BTTS 35%
Group B
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Switzerland7 pts
100%
100% Bosnia and Herzegovina4 pts
100%
0% Canada4 pts
100%
0% Qatar1 pts
0%
0% Canada v Bosnia and Herzegovina
1–1final · predicted 1–0 · ✗ missed
Proj. 1.85–0.68balancedHost (home)
O1.5 71%O2.5 46%O3.5 25%BTTS 41%
Qatar v Switzerland
1–1final · predicted 0–2 · ✗ missed
Proj. 0.80–2.27open, high-scoringNeutral
O1.5 81%O2.5 59%O3.5 37%BTTS 49%
Switzerland v Bosnia and Herzegovina
4–1final · predicted 1–0 · ✓ called it
Proj. 1.80–0.63balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 39%
Canada v Qatar
6–0final · predicted 2–0 · ✓ called it
Proj. 2.34–0.86open, high-scoringHost (home)
O1.5 82%O2.5 62%O3.5 40%BTTS 52%
Bosnia and Herzegovina v Qatar
3–1final · predicted 1–1 · ✗ missed
Proj. 1.11–1.12balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 45%
Switzerland v Canada
2–1final · predicted 1–1 · ✓ called it
Proj. 1.78–1.03balancedHost (home)
O1.5 77%O2.5 53%O3.5 31%BTTS 53%
Group C
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Brazil7 pts
100%
100% Morocco7 pts
100%
0% Haiti0 pts
0%
0% Scotland3 pts
0%
0% Brazil v Morocco
1–1final · predicted 1–0 · ✗ missed
Proj. 1.24–0.60tight, low-scoringNeutral
O1.5 55%O2.5 28%O3.5 12%BTTS 32%
Haiti v Scotland
0–1final · predicted 0–1 · ✓ called it
Proj. 0.92–1.65balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 48%
Scotland v Morocco
0–1final · predicted 0–1 · ✓ called it
Proj. 0.62–1.24tight, low-scoringNeutral
O1.5 55%O2.5 29%O3.5 12%BTTS 32%
Brazil v Haiti
3–0final · predicted 3–0 · ✓ called it
Proj. 3.30–0.45open, high-scoringNeutral
O1.5 88%O2.5 72%O3.5 51%BTTS 34%
Morocco v Haiti
4–2final · predicted 1–0 · ✓ called it
Proj. 1.93–0.53balancedNeutral
O1.5 70%O2.5 44%O3.5 24%BTTS 35%
Scotland v Brazil
0–3final · predicted 0–2 · ✓ called it
Proj. 0.52–2.14balancedNeutral
O1.5 74%O2.5 49%O3.5 28%BTTS 35%
Group D
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
United States6 pts
100%
100% Australia4 pts
100%
0% Paraguay4 pts
100%
0% Turkey3 pts
0%
0% United States v Paraguay
4–1final · predicted 1–0 · ✓ called it
Proj. 1.42–0.84balancedHost (home)
O1.5 66%O2.5 39%O3.5 19%BTTS 43%
Australia v Turkey
2–0final · predicted 1–1 · ✓ called it
Proj. 1.35–1.03balancedNeutral
O1.5 68%O2.5 42%O3.5 22%BTTS 47%
United States v Australia
2–0final · predicted 1–1 · ✓ called it
Proj. 1.29–1.09balancedHost (home)
O1.5 68%O2.5 42%O3.5 22%BTTS 48%
Turkey v Paraguay
0–1final · predicted 1–1 · ✗ missed
Proj. 1.13–1.05balancedNeutral
O1.5 64%O2.5 37%O3.5 18%BTTS 43%
Paraguay v Australia
0–0final · predicted 0–1 · ✗ missed
Proj. 0.75–1.07tight, low-scoringNeutral
O1.5 54%O2.5 27%O3.5 11%BTTS 34%
Turkey v United States
3–2final · predicted 1–1 · ✓ called it
Proj. 1.49–1.40balancedHost (home)
O1.5 78%O2.5 55%O3.5 33%BTTS 58%
Group E
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Germany6 pts
100%
100% Ecuador4 pts
100%
0% Ivory Coast6 pts
100%
0% Curaçao1 pts
0%
0% Germany v Curaçao
7–1final · predicted 3–0 · ✓ called it
Proj. 3.14–0.60open, high-scoringNeutral
O1.5 88%O2.5 72%O3.5 51%BTTS 43%
Ivory Coast v Ecuador
1–0final · predicted 0–1 · ✗ missed
Proj. 0.73–1.04tight, low-scoringNeutral
O1.5 53%O2.5 26%O3.5 11%BTTS 33%
Germany v Ivory Coast
2–1final · predicted 1–0 · ✓ called it
Proj. 1.69–0.93balancedNeutral
O1.5 73%O2.5 49%O3.5 27%BTTS 49%
Ecuador v Curaçao
0–0final · predicted 1–0 · ✗ missed
Proj. 1.94–0.47balancedNeutral
O1.5 69%O2.5 43%O3.5 23%BTTS 32%
Ecuador v Germany
2–1final · predicted 1–1 · ✗ missed
Proj. 1.02–1.30balancedNeutral
O1.5 67%O2.5 41%O3.5 21%BTTS 46%
Curaçao v Ivory Coast
0–2final · predicted 0–1 · ✓ called it
Proj. 0.61–1.77balancedNeutral
O1.5 68%O2.5 42%O3.5 22%BTTS 38%
Group F
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Netherlands7 pts
100%
100% Japan5 pts
100%
0% Sweden4 pts
100%
0% Tunisia0 pts
0%
0% Netherlands v Japan
2–2final · predicted 1–1 · ✗ missed
Proj. 1.48–1.12balancedNeutral
O1.5 73%O2.5 48%O3.5 27%BTTS 52%
Sweden v Tunisia
5–1final · predicted 1–0 · ✓ called it
Proj. 1.15–0.94balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 41%
Netherlands v Sweden
5–1final · predicted 1–0 · ✓ called it
Proj. 1.82–0.97balancedNeutral
O1.5 76%O2.5 53%O3.5 31%BTTS 52%
Tunisia v Japan
0–4final · predicted 0–1 · ✓ called it
Proj. 0.77–1.33balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 39%
Japan v Sweden
1–1final · predicted 1–1 · ✗ missed
Proj. 1.47–1.04balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 49%
Tunisia v Netherlands
1–3final · predicted 0–1 · ✓ called it
Proj. 0.72–1.64balancedNeutral
O1.5 68%O2.5 42%O3.5 21%BTTS 41%
Group G
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Belgium5 pts
100%
100% Egypt5 pts
100%
0% Iran3 pts
0%
0% New Zealand1 pts
0%
0% Belgium v Egypt
1–1final · predicted 1–0 · ✗ missed
Proj. 1.55–0.67balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 38%
Iran v New Zealand
2–2final · predicted 1–0 · ✗ missed
Proj. 1.55–0.70balancedNeutral
O1.5 65%O2.5 39%O3.5 19%BTTS 39%
Belgium v Iran
0–0final · predicted 1–0 · ✗ missed
Proj. 1.47–0.87balancedNeutral
O1.5 68%O2.5 41%O3.5 21%BTTS 44%
New Zealand v Egypt
1–3final · predicted 0–1 · ✓ called it
Proj. 0.73–1.20balancedNeutral
O1.5 57%O2.5 31%O3.5 13%BTTS 36%
Egypt v Iran
1–1final · predicted 0–1 · ✗ missed
Proj. 0.79–1.07tight, low-scoringNeutral
O1.5 55%O2.5 29%O3.5 12%BTTS 35%
New Zealand v Belgium
1–5final · predicted 0–2 · ✓ called it
Proj. 0.60–2.25balancedNeutral
O1.5 77%O2.5 54%O3.5 32%BTTS 40%
Group H
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Spain7 pts
100%
100% Cape Verde3 pts
100%
0% Saudi Arabia2 pts
0%
0% Uruguay2 pts
0%
0% Spain v Cape Verde
0–0final · predicted 2–0 · ✗ missed
Proj. 2.54–0.44open, high-scoringNeutral
O1.5 79%O2.5 57%O3.5 35%BTTS 32%
Saudi Arabia v Uruguay
1–1final · predicted 0–1 · ✗ missed
Proj. 0.48–1.55balancedNeutral
O1.5 60%O2.5 33%O3.5 15%BTTS 30%
Spain v Saudi Arabia
4–0final · predicted 2–0 · ✓ called it
Proj. 2.38–0.45balancedNeutral
O1.5 77%O2.5 54%O3.5 32%BTTS 33%
Uruguay v Cape Verde
2–2final · predicted 1–0 · ✗ missed
Proj. 1.66–0.47balancedNeutral
O1.5 62%O2.5 36%O3.5 17%BTTS 30%
Cape Verde v Saudi Arabia
0–0final · predicted 0–0 · ✗ missed
Proj. 0.84–0.93tight, low-scoringNeutral
O1.5 53%O2.5 26%O3.5 11%BTTS 34%
Uruguay v Spain
0–1final · predicted 0–1 · ✓ called it
Proj. 0.81–1.32balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 40%
Group I
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
France9 pts
100%
100% Norway6 pts
100%
0% Senegal3 pts
100%
0% Iraq0 pts
0%
0% France v Senegal
3–1final · predicted 1–0 · ✓ called it
Proj. 1.45–0.76balancedNeutral
O1.5 64%O2.5 38%O3.5 18%BTTS 40%
Iraq v Norway
1–4final · predicted 0–1 · ✓ called it
Proj. 0.66–1.61balancedNeutral
O1.5 66%O2.5 39%O3.5 19%BTTS 38%
France v Iraq
3–0final · predicted 2–0 · ✓ called it
Proj. 2.06–0.49balancedNeutral
O1.5 72%O2.5 47%O3.5 25%BTTS 34%
Norway v Senegal
3–2final · predicted 1–0 · ✓ called it
Proj. 1.13–1.01balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 43%
Norway v France
1–4final · predicted 0–1 · ✓ called it
Proj. 0.95–1.62balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 49%
Senegal v Iraq
5–0final · predicted 1–0 · ✓ called it
Proj. 1.28–0.59tight, low-scoringNeutral
O1.5 56%O2.5 29%O3.5 12%BTTS 32%
Group J
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Argentina9 pts
100%
100% Algeria4 pts
100%
0% Austria4 pts
100%
0% Jordan0 pts
0%
0% Argentina v Algeria
3–0final · predicted 1–0 · ✓ called it
Proj. 1.80–0.70balancedNeutral
O1.5 71%O2.5 45%O3.5 24%BTTS 42%
Austria v Jordan
3–1final · predicted 1–0 · ✓ called it
Proj. 1.78–0.74balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 43%
Argentina v Austria
2–0final · predicted 1–0 · ✓ called it
Proj. 1.64–0.66balancedNeutral
O1.5 67%O2.5 40%O3.5 20%BTTS 39%
Jordan v Algeria
1–2final · predicted 0–1 · ✓ called it
Proj. 0.81–1.89balancedNeutral
O1.5 75%O2.5 50%O3.5 29%BTTS 47%
Jordan v Argentina
1–3final · predicted 0–2 · ✓ called it
Proj. 0.43–2.61open, high-scoringNeutral
O1.5 80%O2.5 58%O3.5 36%BTTS 32%
Algeria v Austria
3–3final · predicted 1–1 · ✗ missed
Proj. 1.19–1.23balancedNeutral
O1.5 69%O2.5 43%O3.5 23%BTTS 49%
Group K
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
Colombia7 pts
100%
100% DR Congo4 pts
100%
0% Portugal5 pts
100%
0% Uzbekistan0 pts
0%
0% Portugal v DR Congo
1–1final · predicted 1–0 · ✗ missed
Proj. 1.94–0.49balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 33%
Uzbekistan v Colombia
1–3final · predicted 0–1 · ✓ called it
Proj. 0.57–1.72balancedNeutral
O1.5 66%O2.5 40%O3.5 20%BTTS 36%
Portugal v Uzbekistan
5–0final · predicted 1–0 · ✓ called it
Proj. 1.92–0.60balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 38%
Colombia v DR Congo
1–0final · predicted 1–0 · ✓ called it
Proj. 1.73–0.47balancedNeutral
O1.5 64%O2.5 38%O3.5 18%BTTS 31%
DR Congo v Uzbekistan
3–1final · predicted 0–0 · ✗ missed
Proj. 0.77–0.95tight, low-scoringNeutral
O1.5 51%O2.5 25%O3.5 10%BTTS 32%
Colombia v Portugal
0–0final · predicted 1–1 · ✗ missed
Proj. 1.09–1.17balancedNeutral
O1.5 66%O2.5 39%O3.5 20%BTTS 45%
Group L
▾Chance to reach the knockouts6/6 played · top 2 + 8 best 3rd-placed advance
advwin
England7 pts
100%
100% Croatia6 pts
100%
0% Ghana4 pts
100%
0% Panama0 pts
0%
0% England v Croatia
4–2final · predicted 1–0 · ✓ called it
Proj. 1.57–0.83balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 44%
Ghana v Panama
1–0final · predicted 1–1 · ✓ called it
Proj. 1.11–1.04balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 43%
England v Ghana
0–0final · predicted 2–0 · ✗ missed
Proj. 2.11–0.45balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 32%
Panama v Croatia
0–1final · predicted 0–1 · ✓ called it
Proj. 0.77–2.02balancedNeutral
O1.5 77%O2.5 53%O3.5 31%BTTS 46%
Croatia v Ghana
2–1final · predicted 1–0 · ✓ called it
Proj. 1.72–0.70balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 41%
Panama v England
0–2final · predicted 0–2 · ✓ called it
Proj. 0.50–2.48open, high-scoringNeutral
O1.5 79%O2.5 57%O3.5 35%BTTS 36%
Upcoming matches
chronological · soonest firstNothing here yet.
Results
chronological · every played matchThu · Jun 11
Mexico v South Africa
2–0final · predicted 1–0 · ✓ called it
Proj. 1.75–0.62balancedHost (home)
O1.5 68%O2.5 42%O3.5 22%BTTS 38%
South Korea v Czech Republic
2–1final · predicted 1–1 · ✓ called it
Proj. 1.26–1.13balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 48%
Fri · Jun 12
Canada v Bosnia and Herzegovina
1–1final · predicted 1–0 · ✗ missed
Proj. 1.85–0.68balancedHost (home)
O1.5 71%O2.5 46%O3.5 25%BTTS 41%
United States v Paraguay
4–1final · predicted 1–0 · ✓ called it
Proj. 1.42–0.84balancedHost (home)
O1.5 66%O2.5 39%O3.5 19%BTTS 43%
Sat · Jun 13
Qatar v Switzerland
1–1final · predicted 0–2 · ✗ missed
Proj. 0.80–2.27open, high-scoringNeutral
O1.5 81%O2.5 59%O3.5 37%BTTS 49%
Brazil v Morocco
1–1final · predicted 1–0 · ✗ missed
Proj. 1.24–0.60tight, low-scoringNeutral
O1.5 55%O2.5 28%O3.5 12%BTTS 32%
Haiti v Scotland
0–1final · predicted 0–1 · ✓ called it
Proj. 0.92–1.65balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 48%
Sun · Jun 14
Australia v Turkey
2–0final · predicted 1–1 · ✓ called it
Proj. 1.35–1.03balancedNeutral
O1.5 68%O2.5 42%O3.5 22%BTTS 47%
Germany v Curaçao
7–1final · predicted 3–0 · ✓ called it
Proj. 3.14–0.60open, high-scoringNeutral
O1.5 88%O2.5 72%O3.5 51%BTTS 43%
Netherlands v Japan
2–2final · predicted 1–1 · ✗ missed
Proj. 1.48–1.12balancedNeutral
O1.5 73%O2.5 48%O3.5 27%BTTS 52%
Ivory Coast v Ecuador
1–0final · predicted 0–1 · ✗ missed
Proj. 0.73–1.04tight, low-scoringNeutral
O1.5 53%O2.5 26%O3.5 11%BTTS 33%
Sweden v Tunisia
5–1final · predicted 1–0 · ✓ called it
Proj. 1.15–0.94balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 41%
Mon · Jun 15
Spain v Cape Verde
0–0final · predicted 2–0 · ✗ missed
Proj. 2.54–0.44open, high-scoringNeutral
O1.5 79%O2.5 57%O3.5 35%BTTS 32%
Belgium v Egypt
1–1final · predicted 1–0 · ✗ missed
Proj. 1.55–0.67balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 38%
Saudi Arabia v Uruguay
1–1final · predicted 0–1 · ✗ missed
Proj. 0.48–1.55balancedNeutral
O1.5 60%O2.5 33%O3.5 15%BTTS 30%
Iran v New Zealand
2–2final · predicted 1–0 · ✗ missed
Proj. 1.55–0.70balancedNeutral
O1.5 65%O2.5 39%O3.5 19%BTTS 39%
Tue · Jun 16
France v Senegal
3–1final · predicted 1–0 · ✓ called it
Proj. 1.45–0.76balancedNeutral
O1.5 64%O2.5 38%O3.5 18%BTTS 40%
Iraq v Norway
1–4final · predicted 0–1 · ✓ called it
Proj. 0.66–1.61balancedNeutral
O1.5 66%O2.5 39%O3.5 19%BTTS 38%
Argentina v Algeria
3–0final · predicted 1–0 · ✓ called it
Proj. 1.80–0.70balancedNeutral
O1.5 71%O2.5 45%O3.5 24%BTTS 42%
Wed · Jun 17
Austria v Jordan
3–1final · predicted 1–0 · ✓ called it
Proj. 1.78–0.74balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 43%
Portugal v DR Congo
1–1final · predicted 1–0 · ✗ missed
Proj. 1.94–0.49balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 33%
England v Croatia
4–2final · predicted 1–0 · ✓ called it
Proj. 1.57–0.83balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 44%
Ghana v Panama
1–0final · predicted 1–1 · ✓ called it
Proj. 1.11–1.04balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 43%
Uzbekistan v Colombia
1–3final · predicted 0–1 · ✓ called it
Proj. 0.57–1.72balancedNeutral
O1.5 66%O2.5 40%O3.5 20%BTTS 36%
Thu · Jun 18
Czech Republic v South Africa
1–1final · predicted 1–0 · ✗ missed
Proj. 1.30–0.81balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 40%
Switzerland v Bosnia and Herzegovina
4–1final · predicted 1–0 · ✓ called it
Proj. 1.80–0.63balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 39%
Canada v Qatar
6–0final · predicted 2–0 · ✓ called it
Proj. 2.34–0.86open, high-scoringHost (home)
O1.5 82%O2.5 62%O3.5 40%BTTS 52%
Mexico v South Korea
1–0final · predicted 1–0 · ✓ called it
Proj. 1.52–0.96balancedHost (home)
O1.5 71%O2.5 45%O3.5 24%BTTS 48%
Fri · Jun 19
United States v Australia
2–0final · predicted 1–1 · ✓ called it
Proj. 1.29–1.09balancedHost (home)
O1.5 68%O2.5 42%O3.5 22%BTTS 48%
Scotland v Morocco
0–1final · predicted 0–1 · ✓ called it
Proj. 0.62–1.24tight, low-scoringNeutral
O1.5 55%O2.5 29%O3.5 12%BTTS 32%
Brazil v Haiti
3–0final · predicted 3–0 · ✓ called it
Proj. 3.30–0.45open, high-scoringNeutral
O1.5 88%O2.5 72%O3.5 51%BTTS 34%
Turkey v Paraguay
0–1final · predicted 1–1 · ✗ missed
Proj. 1.13–1.05balancedNeutral
O1.5 64%O2.5 37%O3.5 18%BTTS 43%
Sat · Jun 20
Netherlands v Sweden
5–1final · predicted 1–0 · ✓ called it
Proj. 1.82–0.97balancedNeutral
O1.5 76%O2.5 53%O3.5 31%BTTS 52%
Germany v Ivory Coast
2–1final · predicted 1–0 · ✓ called it
Proj. 1.69–0.93balancedNeutral
O1.5 73%O2.5 49%O3.5 27%BTTS 49%
Ecuador v Curaçao
0–0final · predicted 1–0 · ✗ missed
Proj. 1.94–0.47balancedNeutral
O1.5 69%O2.5 43%O3.5 23%BTTS 32%
Sun · Jun 21
Tunisia v Japan
0–4final · predicted 0–1 · ✓ called it
Proj. 0.77–1.33balancedNeutral
O1.5 62%O2.5 35%O3.5 16%BTTS 39%
Spain v Saudi Arabia
4–0final · predicted 2–0 · ✓ called it
Proj. 2.38–0.45balancedNeutral
O1.5 77%O2.5 54%O3.5 32%BTTS 33%
Belgium v Iran
0–0final · predicted 1–0 · ✗ missed
Proj. 1.47–0.87balancedNeutral
O1.5 68%O2.5 41%O3.5 21%BTTS 44%
Uruguay v Cape Verde
2–2final · predicted 1–0 · ✗ missed
Proj. 1.66–0.47balancedNeutral
O1.5 62%O2.5 36%O3.5 17%BTTS 30%
New Zealand v Egypt
1–3final · predicted 0–1 · ✓ called it
Proj. 0.73–1.20balancedNeutral
O1.5 57%O2.5 31%O3.5 13%BTTS 36%
Mon · Jun 22
Argentina v Austria
2–0final · predicted 1–0 · ✓ called it
Proj. 1.64–0.66balancedNeutral
O1.5 67%O2.5 40%O3.5 20%BTTS 39%
France v Iraq
3–0final · predicted 2–0 · ✓ called it
Proj. 2.06–0.49balancedNeutral
O1.5 72%O2.5 47%O3.5 25%BTTS 34%
Norway v Senegal
3–2final · predicted 1–0 · ✓ called it
Proj. 1.13–1.01balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 43%
Jordan v Algeria
1–2final · predicted 0–1 · ✓ called it
Proj. 0.81–1.89balancedNeutral
O1.5 75%O2.5 50%O3.5 29%BTTS 47%
Tue · Jun 23
Portugal v Uzbekistan
5–0final · predicted 1–0 · ✓ called it
Proj. 1.92–0.60balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 38%
England v Ghana
0–0final · predicted 2–0 · ✗ missed
Proj. 2.11–0.45balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 32%
Panama v Croatia
0–1final · predicted 0–1 · ✓ called it
Proj. 0.77–2.02balancedNeutral
O1.5 77%O2.5 53%O3.5 31%BTTS 46%
Colombia v DR Congo
1–0final · predicted 1–0 · ✓ called it
Proj. 1.73–0.47balancedNeutral
O1.5 64%O2.5 38%O3.5 18%BTTS 31%
Wed · Jun 24
Bosnia and Herzegovina v Qatar
3–1final · predicted 1–1 · ✗ missed
Proj. 1.11–1.12balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 45%
Switzerland v Canada
2–1final · predicted 1–1 · ✓ called it
Proj. 1.78–1.03balancedHost (home)
O1.5 77%O2.5 53%O3.5 31%BTTS 53%
Morocco v Haiti
4–2final · predicted 1–0 · ✓ called it
Proj. 1.93–0.53balancedNeutral
O1.5 70%O2.5 44%O3.5 24%BTTS 35%
Scotland v Brazil
0–3final · predicted 0–2 · ✓ called it
Proj. 0.52–2.14balancedNeutral
O1.5 74%O2.5 49%O3.5 28%BTTS 35%
Czech Republic v Mexico
0–3final · predicted 1–1 · ✓ called it
Proj. 1.29–1.37balancedHost (home)
O1.5 74%O2.5 49%O3.5 28%BTTS 54%
South Africa v South Korea
1–0final · predicted 0–1 · ✗ missed
Proj. 0.70–1.25balancedNeutral
O1.5 58%O2.5 31%O3.5 13%BTTS 35%
Thu · Jun 25
Ecuador v Germany
2–1final · predicted 1–1 · ✗ missed
Proj. 1.02–1.30balancedNeutral
O1.5 67%O2.5 41%O3.5 21%BTTS 46%
Curaçao v Ivory Coast
0–2final · predicted 0–1 · ✓ called it
Proj. 0.61–1.77balancedNeutral
O1.5 68%O2.5 42%O3.5 22%BTTS 38%
Japan v Sweden
1–1final · predicted 1–1 · ✗ missed
Proj. 1.47–1.04balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 49%
Tunisia v Netherlands
1–3final · predicted 0–1 · ✓ called it
Proj. 0.72–1.64balancedNeutral
O1.5 68%O2.5 42%O3.5 21%BTTS 41%
Paraguay v Australia
0–0final · predicted 0–1 · ✗ missed
Proj. 0.75–1.07tight, low-scoringNeutral
O1.5 54%O2.5 27%O3.5 11%BTTS 34%
Turkey v United States
3–2final · predicted 1–1 · ✓ called it
Proj. 1.49–1.40balancedHost (home)
O1.5 78%O2.5 55%O3.5 33%BTTS 58%
Fri · Jun 26
Norway v France
1–4final · predicted 0–1 · ✓ called it
Proj. 0.95–1.62balancedNeutral
O1.5 72%O2.5 47%O3.5 26%BTTS 49%
Senegal v Iraq
5–0final · predicted 1–0 · ✓ called it
Proj. 1.28–0.59tight, low-scoringNeutral
O1.5 56%O2.5 29%O3.5 12%BTTS 32%
Cape Verde v Saudi Arabia
0–0final · predicted 0–0 · ✗ missed
Proj. 0.84–0.93tight, low-scoringNeutral
O1.5 53%O2.5 26%O3.5 11%BTTS 34%
Uruguay v Spain
0–1final · predicted 0–1 · ✓ called it
Proj. 0.81–1.32balancedNeutral
O1.5 63%O2.5 36%O3.5 17%BTTS 40%
Egypt v Iran
1–1final · predicted 0–1 · ✗ missed
Proj. 0.79–1.07tight, low-scoringNeutral
O1.5 55%O2.5 29%O3.5 12%BTTS 35%
New Zealand v Belgium
1–5final · predicted 0–2 · ✓ called it
Proj. 0.60–2.25balancedNeutral
O1.5 77%O2.5 54%O3.5 32%BTTS 40%
Sat · Jun 27
Croatia v Ghana
2–1final · predicted 1–0 · ✓ called it
Proj. 1.72–0.70balancedNeutral
O1.5 69%O2.5 44%O3.5 23%BTTS 41%
Panama v England
0–2final · predicted 0–2 · ✓ called it
Proj. 0.50–2.48open, high-scoringNeutral
O1.5 79%O2.5 57%O3.5 35%BTTS 36%
DR Congo v Uzbekistan
3–1final · predicted 0–0 · ✗ missed
Proj. 0.77–0.95tight, low-scoringNeutral
O1.5 51%O2.5 25%O3.5 10%BTTS 32%
Colombia v Portugal
0–0final · predicted 1–1 · ✗ missed
Proj. 1.09–1.17balancedNeutral
O1.5 66%O2.5 39%O3.5 20%BTTS 45%
Jordan v Argentina
1–3final · predicted 0–2 · ✓ called it
Proj. 0.43–2.61open, high-scoringNeutral
O1.5 80%O2.5 58%O3.5 36%BTTS 32%
Algeria v Austria
3–3final · predicted 1–1 · ✗ missed
Proj. 1.19–1.23balancedNeutral
O1.5 69%O2.5 43%O3.5 23%BTTS 49%
Sun · Jun 28
South Africa v Canada
0–1final · predicted 0–1 · ✓ called it
Proj. 0.85–1.34balancedHost (home)
O1.5 64%O2.5 37%O3.5 18%BTTS 42%
Mon · Jun 29
Brazil v Japan
2–1final · predicted 1–0 · ✓ called it
Proj. 1.64–0.76balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 42%
Germany v Paraguay
1–1final · predicted 1–0 · ✗ missed
Proj. 1.64–0.83balancedNeutral
O1.5 70%O2.5 45%O3.5 24%BTTS 45%
Netherlands v Morocco
1–1final · predicted 1–0 · ✗ missed
Proj. 1.12–0.89balancedNeutral
O1.5 59%O2.5 33%O3.5 15%BTTS 39%
Tue · Jun 30
Ivory Coast v Norway
1–2final · predicted 0–1 · ✓ called it
Proj. 0.93–1.30balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 43%
France v Sweden
3–0final · predicted 1–0 · ✓ called it
Proj. 1.82–0.85balancedNeutral
O1.5 74%O2.5 50%O3.5 28%BTTS 48%
Mexico v Ecuador
2–0final · predicted 1–0 · ✓ called it
Proj. 1.11–0.97balancedHost (home)
O1.5 61%O2.5 35%O3.5 16%BTTS 41%
Wed · Jul 01
England v DR Congo
2–1final · predicted 1–0 · ✓ called it
Proj. 1.88–0.42balancedNeutral
O1.5 66%O2.5 40%O3.5 20%BTTS 29%
Belgium v Senegal
3–2final · predicted 1–0 · ✓ called it
Proj. 1.44–0.79balancedNeutral
O1.5 65%O2.5 39%O3.5 19%BTTS 41%
United States v Bosnia and Herzegovina
2–0final · predicted 1–0 · ✓ called it
Proj. 1.90–0.67balancedHost (home)
O1.5 72%O2.5 47%O3.5 26%BTTS 41%
Thu · Jul 02
Spain v Austria
3–0final · predicted 1–0 · ✓ called it
Proj. 1.81–0.77balancedNeutral
O1.5 73%O2.5 47%O3.5 26%BTTS 44%
Portugal v Croatia
2–1final · predicted 1–0 · ✓ called it
Proj. 1.62–0.97balancedNeutral
O1.5 73%O2.5 48%O3.5 26%BTTS 49%
Switzerland v Algeria
2–0final · predicted 1–1 · ✓ called it
Proj. 1.44–1.16balancedNeutral
O1.5 73%O2.5 48%O3.5 27%BTTS 52%
Fri · Jul 03
Australia v Egypt
1–1final · predicted 1–0 · ✗ missed
Proj. 1.04–0.78tight, low-scoringNeutral
O1.5 54%O2.5 27%O3.5 11%BTTS 34%
Argentina v Cape Verde
3–2final · predicted 2–0 · ✓ called it
Proj. 2.30–0.38balancedNeutral
O1.5 74%O2.5 50%O3.5 28%BTTS 28%
Colombia v Ghana
1–0final · predicted 1–0 · ✓ called it
Proj. 1.95–0.51balancedNeutral
O1.5 70%O2.5 44%O3.5 23%BTTS 34%
Sat · Jul 04
Canada v Morocco
0–3final · predicted 0–1 · ✓ called it
Proj. 0.97–1.06balancedHost (home)
O1.5 60%O2.5 33%O3.5 15%BTTS 40%
Paraguay v France
0–1final · predicted 0–1 · ✓ called it
Proj. 0.63–1.60balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 37%
Sun · Jul 05
Brazil v Norway
1–2final · predicted 1–0 · ✗ missed
Proj. 1.79–0.73balancedNeutral
O1.5 71%O2.5 46%O3.5 25%BTTS 43%
Mexico v England
2–3final · predicted 0–1 · ✓ called it
Proj. 0.91–1.51balancedHost (home)
O1.5 69%O2.5 43%O3.5 23%BTTS 46%
Mon · Jul 06
Portugal v Spain
0–1final · predicted 1–1 · ✓ called it
Proj. 1.09–1.39balancedNeutral
O1.5 71%O2.5 45%O3.5 24%BTTS 49%
United States v Belgium
1–4final · predicted 1–1 · ✓ called it
Proj. 1.12–1.63balancedHost (home)
O1.5 76%O2.5 52%O3.5 30%BTTS 54%
Tue · Jul 07
Argentina v Egypt
3–2final · predicted 1–0 · ✓ called it
Proj. 1.63–0.49balancedNeutral
O1.5 62%O2.5 35%O3.5 17%BTTS 31%
Switzerland v Colombia
0–0final · predicted 0–1 · ✗ missed
Proj. 0.97–1.39balancedNeutral
O1.5 68%O2.5 42%O3.5 21%BTTS 46%
Thu · Jul 09
France v Morocco
2–0final · predicted 1–0 · ✓ called it
Proj. 1.12–0.78balancedNeutral
O1.5 57%O2.5 30%O3.5 13%BTTS 36%
Fri · Jul 10
Spain v Belgium
2–1final · predicted 1–1 · ✓ called it
Proj. 1.46–1.07balancedNeutral
O1.5 72%O2.5 46%O3.5 25%BTTS 50%
Sat · Jul 11
Norway v England
1–2final · predicted 0–1 · ✓ called it
Proj. 0.81–1.59balancedNeutral
O1.5 69%O2.5 43%O3.5 22%BTTS 44%
Argentina v Switzerland
3–1final · predicted 1–0 · ✓ called it
Proj. 1.60–0.77balancedNeutral
O1.5 68%O2.5 42%O3.5 22%BTTS 43%
Tue · Jul 14
France v Spain
0–2final · predicted 1–1 · ✓ called it
Proj. 1.08–1.40balancedNeutral
O1.5 71%O2.5 45%O3.5 24%BTTS 49%
Wed · Jul 15
England v Argentina
1–2final · predicted 0–1 · ✓ called it
Proj. 0.91–1.08balancedNeutral
O1.5 59%O2.5 32%O3.5 14%BTTS 39%
Sat · Jul 18
France v England
4–6final · predicted 1–1 · ✓ called it
Proj. 1.04–1.19balancedNeutral
O1.5 65%O2.5 38%O3.5 19%BTTS 45%
Sun · Jul 19
Spain v Argentina
1–0final · predicted 1–1 · ✗ missed
Proj. 1.06–1.13balancedNeutral
O1.5 64%O2.5 37%O3.5 18%BTTS 44%
Knockout bracket
The verified WC2026 knockout tree. Empty slots show the model's projected qualifier (top two per group by chance to advance); Best 3rd-placed slots are cross-group and filled once the eight qualifying third places are known.
Round of 32
South Africa38%
Canada62%
0–1
Brazil70%
Japan30%
2–1
Germany69%
Paraguay31%
1–1
Netherlands56%
Morocco44%
1–1
Ivory Coast41%
Norway59%
1–2
France71%
Sweden29%
3–0
Mexico54%
Ecuador46%
2–0
England82%
DR Congo18%
2–1
Belgium66%
Senegal34%
3–2
United States77%
Bosnia and Herzegovina23%
2–0
Spain73%
Austria27%
3–0
Portugal65%
Croatia35%
2–1
Switzerland56%
Algeria44%
2–0
Australia57%
Egypt43%
1–1
Argentina88%
Cape Verde12%
3–2
Colombia81%
Ghana19%
1–0
Round of 16
Canada48%
Morocco52%
0–3
Paraguay27%
France73%
0–1
Brazil74%
Norway26%
1–2
Mexico36%
England64%
2–3
Portugal43%
Spain57%
0–1
United States39%
Belgium61%
1–4
Argentina77%
Egypt23%
3–2
Switzerland40%
Colombia60%
0–0
Quarter-finals
France59%
Morocco41%
2–0
Spain59%
Belgium41%
2–1
Norway32%
England68%
1–2
Argentina69%
Switzerland31%
3–1
Semi-finals
France42%
Spain58%
0–2
England45%
Argentina55%
1–2
Final
Spain48%
Argentina52%
1–0
Best Bets
Only named edges appear here. Curated from the model vs the market plus the external signals the goals-only model is blind to (injuries, rotation, penalty-shootout profiles). Divergence is not value — a bet earns a slot only when we can name why the market is wrong. As of 2026-07-17 · Final + 3rd place. Informational, not betting advice.
No open bets — the tournament is down to two games and both markets are efficient. FINAL, Spain v Argentina: the model leans Argentina (52% to lift the cup, on the Dibu Martínez shootout edge + reigning-champ pedigree), but Polymarket favours SPAIN 59% — and that's the defensible number: Spain have been the tournament's best side (6+ clean sheets, beat France 2-0 in the semi), which a goals-only model under-weights. The model's Argentina lean is most likely its blind spot, not an edge → Pass (Argentina's shootout dominance keeps the trophy genuinely live in the ~29% of finals that reach penalties). 3RD PLACE, France v England: the model likes England (40% vs France's ~50% market), but it's a low-stakes bronze game → Pass. ✓ Tournament track record: R16 calls England-to-advance (Mexico 2-3 England) and Belgium-to-advance (USA 1-4 Belgium) both WON; the QF/SF slates were correctly all-Pass.
Penalty-shootout scoutkeeper style · shooters · who wins a shootout
A 90-minute draw goes to extra time, then penalties. Our model assumes a naive 50/50 shootout; this replaces it with researched numbers. Keeper style: pre-committer dives early off a scouted dossier, reactor waits and reads the shot, mixed adds deception, unknown has no public profile. Penalties decide only ~3–8% of ties, so this is a tiebreaker — except where a keeper is elite (e.g. Bono).
Spain v Argentina Jul 19 · FINAL
Shootout Spain 30% / Argentina 70% · draw90 29% · edge Argentina (huge)
🧤 Spain: Unai Simón reactor
🧤 Argentina: Emi Martínez mixed
Spain: Oyarzábal 87.5%, Rodri, YamalArgentina: Messi 84.6% (shootout), Lautaro, Álvarez, Dybala
The decisive sub-plot of the final. Dibu Martínez is the world's best shootout GK (Argentina 6-1 all-time, won the '22 title on penalties); Spain have the WORST WC shootout record of any nation (1W-4L). Polymarket favours Spain 59% to lift the cup (they've been the better side in open play), but in the ~29% of finals that reach penalties Argentina are heavy favourites — which is why the model (Argentina 52% incl. pens) and the reigning-champ narrative keep the trophy genuinely live.
France v England Jul 18 · 3rd place
Shootout France 45% / England 55% · draw90 29% · edge England (slight)
🧤 France: Maignan reactor
🧤 England: Pickford pre-committer
France: Mbappé, Olise, DembéléEngland: Kane 88%, Saka, Rashford, Eze
England are 1W-3L but improving (Pickford's dossier + a deep taker pool); France are 2W-3L with back-to-back final losses ('06, '22). Slight England edge if it goes the distance — but this is a low-stakes bronze game where motivation, not pens, decides it.
Model vs Market
Live market comparison is currently unavailable (no open Polymarket markets resolved at build time). It returns automatically once upcoming-match markets are live.
Live Fair Value
No upcoming matches to track yet.
How this works
Every number on this page comes from one Bayesian statistical model fit to historical results — no hand-tuning, no bookmaker odds. Here's the honest version of what it does and what it can't.
The model
A Bayesian hierarchical double-Poisson (the Dixon–Coles lineage). Each team carries a latent attack and defense strength; goals are Poisson-distributed around those strengths, with a single home-advantage term that switches off at neutral venues.
log λ_home = intercept + home_adv·(1−neutral) + atk[home] − def[away] log λ_away = intercept + atk[away] − def[home]
It's fit by MCMC (PyMC + nutpie), so every prediction integrates over the full posterior — the scoreline grid reflects genuine parameter uncertainty, not a single point estimate.
The data
The CC0 martj42/international_results dataset — 8,000+ men's international matches since 2018, 284 teams. The model uses only goals, venue (neutral flag), date and tournament.
It deliberately has no xG, shots, lineups, player ratings, or betting odds — those barely exist for international football and aren't in a free, redistributable source. This is a goals-only model, and honest about it.
Does it actually work?
On a temporal backtest (train on the past, predict the future, no leakage) it scores RPS 0.166 — about 9% better than Elo (0.182). RPS is the standard proper score for ordered 1X2 outcomes.
It's also well-calibrated — when it says 30%, it happens about 30% of the time. The reliability curve should hug the diagonal:
Reliability across all home/draw/away probabilities on 8152 matches (24456 probability–outcome pairs). Points on the diagonal = perfectly calibrated; point size = bin count. Mean miscalibration (ECE) 1.8%. In-sample check — the backtest scores are the out-of-sample skill measure.
On the 104 WC2026 matches played so far — a small, noisy sample — it has called 51% of over/under-2.5 totals correctly and nailed 13/104 exact final scores, placing on average 17% of its probability on the actual goal total. With only ~104 matches these in-tournament figures will bounce around; the backtest numbers above are the stable measure of skill.
| Model | RPS | log-loss | Brier | acc |
|---|---|---|---|---|
| Double-Poisson (shipped) | 0.1659 | 0.861 | 0.505 | 60.1% |
| + Dixon–Coles | 0.1659 | 0.860 | 0.505 | 60.1% |
| + Confederation layer ★ | 0.1651 | 0.857 | 0.504 | 59.8% |
| + Both ★ | 0.1651 | 0.856 | 0.504 | 60.1% |
| Elo baseline | 0.1818 | 0.937 | 0.547 | 49.3% |
Lower RPS / log-loss / Brier is better. ★ = best. Expanding-window temporal backtest, no leakage.
What we tested and rejected
The interesting part of a model is what didn't help. On the same backtest, all of these were within run-to-run noise or worse, so they're off:
- Time-decay weighting (recent matches count more) — best variant ~−0.3% RPS, inside the noise.
- Tournament-importance weighting — slightly hurt.
- Blending with Elo — a weighted average only got worse as more Elo was added; pure Bayesian was optimal.
- Confederation hierarchy (shrink teams toward their AFC/CAF/… mean) — a hair better on paper but inside the noise, and it muddied the model's convergence, so it's off.
- Dixon–Coles low-score correction — once correctly bounded, it's RPS-neutral (a genuinely tiny effect for international football).
The model sits at the goals-only information ceiling: with this data, the plain double-Poisson is as good as it gets. We measured that rather than assuming it.
Source: martj42/international_results (CC0). Built with Python, pandas, PyMC. Predictions are probabilistic — a 70% favourite still loses three times in ten.