Divergence Log
The site's namesake: every completed match, scored against all three sources. Where did the statistical model, the sportsbook consensus, and Polymarket agree — and where did they diverge? For each result we show what each source put on the outcome that actually happened, the model-vs-market gap, and which source called it.
Who's been closer — attribution
Early days — rates and scores will stabilise as more matches are played (n shown per source below).
| Source | Distinct verdict wins | Win rate | Brier | Log loss | Scored on |
|---|---|---|---|---|---|
| Model | 26 | 25% | 0.525 | 0.869 | n = 104 |
| Sportsbook | 0 | 0% | 0.498 | 0.844 | n = 104 |
| Polymarket | 5 | 5% | 0.498 | 0.842 | n = 104 |
Over 104 scored matches: no-credit verdicts — markets agreed 6 · sources agreed 32 · all missed 35.
A "distinct verdict win" means that source put a clearly higher probability (> 2pp) on the actual outcome than every other source. When the two markets jointly beat the model they share a no-credit "markets agreed" verdict — neither book nor Polymarket is credited individually, so a market's win count never double-counts. Brier and log loss score every source on every match where it posted a price (lower is better), independent of the verdict. See the methodology and calibration page.
Performance by divergence size
The site's central question, made measurable: when the model and the market disagree most, who is right? Each row groups completed matches by the size of the model-vs-sportsbook gap on the outcome that actually happened. Brier and log loss score every source that posted a price; lower is better.
| Model–market gap | Matches | Model verdict wins | Model Brier | Sportsbook Brier | Polymarket Brier |
|---|---|---|---|---|---|
| Under 5pp | 17 | 0 | 0.645 n=17 | 0.626 n=17 | 0.632 n=17 |
| 5–15pp | 53 | 19 | 0.483 n=53 | 0.488 n=53 | 0.485 n=53 |
| 15pp+ | 34 | 7 | 0.530 n=34 | 0.450 n=34 | 0.451 n=34 |
"Verdict wins" counts only matches where the model put a clearly higher probability (> 2pp) on the actual outcome than every other source. The largest-gap bucket is where the model most backs itself against the market — the honest test of whether those disagreements pay off.
Rolling calibration by source
Cumulative Brier score per source, from match 5 onward — the first four are omitted because a running average over 1–4 results is noise-dominated. Each market's line advances only on matches it priced. Lower is better — a line drifting up over time means calibration is getting worse.
Model
Sportsbook
Polymarket
Where the model diverged most — and how it turned out
The biggest model-vs-market gaps (15pp+), split by result. Percentages are each source's frozen pre-match probability on the outcome that actually happened. For the full reliability picture across all matches, see the calibration page.
Biggest calls that came in (showing top 3 of 7)
- Croatia v Ghana 2–1 Home win Model 78%Sportsbook 54%Polymarket 54%
- Mexico v Czechia 3–0 Home win Model 71%Sportsbook 50%Polymarket 51%
- Colombia v Ghana 1–0 Home win Model 86%Sportsbook 67%Polymarket 68%
Biggest calls that missed (showing top 3 of 15)
- Ghana v Panama 1–0 Home win Model 11%Sportsbook 44%Polymarket 44%
- DR Congo v Uzbekistan 3–1 Home win Model 24%Sportsbook 56%Polymarket 56%
- Belgium v Iran 0–0 Draw Model 32%Sportsbook 21%Polymarket 21%
Match archive
| Date | Match | Stage | Result | Model | Sportsbook | Polymarket | Gap | Divergence | Verdict |
|---|---|---|---|---|---|---|---|---|---|
| 19 July 2026 | Spain v Argentina | Final | 1–0 Draw | 31% | 32% | 31% | 1pp | Market more confident than the model | All missed |
| 18 July 2026 | France v England | Third place | 4–6 Away win | 27% | 25% | 25% | 10pp | Market more confident than the model | All missed |
| 15 July 2026 | England v Argentina | Semi-final | 1–2 Away win | 48% | 32% | 32% | 17pp | Model and market disagree on the favourite | Model closest |
| 14 July 2026 | France v Spain | Semi-final | 0–2 Away win | 48% | 31% | 30% | 17pp | Model and market disagree on the favourite | Model closest |
| 11 July 2026 | Norway v England | Quarter-final | 1–2 Draw | 30% | 26% | 26% | 4pp | Market more confident than the model | All missed |
| 11 July 2026 | Argentina v Switzerland | Quarter-final | 3–1 Draw | 23% | 27% | 27% | 13pp | Market hedges more than the model | All missed |
| 10 July 2026 | Spain v Belgium | Quarter-final | 2–1 Home win | 75% | 59% | 59% | 17pp | Market hedges more than the model | Model closest |
| 9 July 2026 | France v Morocco | Quarter-final | 2–0 Home win | 49% | 61% | 62% | 11pp | Market more confident than the model | Sources agreed |
| 7 July 2026 | Argentina v Egypt | Round of 16 | 3–2 Home win | 84% | 71% | 72% | 12pp | Market hedges more than the model | Model closest |
| 7 July 2026 | Switzerland v Colombia | Round of 16 | 0–0 Draw | 30% | 31% | 31% | 6pp | Market hedges more than the model | All missed |
| 6 July 2026 | Portugal v Spain | Round of 16 | 0–1 Away win | 55% | 50% | 51% | 7pp | Market hedges more than the model | Model closest |
| 6 July 2026 | USA v Belgium | Round of 16 | 1–4 Away win | 44% | 34% | 34% | 13pp | Model and market disagree on the favourite | Model closest |
| 5 July 2026 | Brazil v Norway | Round of 16 | 1–2 Away win | 23% | 21% | 19% | 7pp | Market more confident than the model | All missed |
| 5 July 2026 | Mexico v England | Round of 16 | 2–3 Away win | 45% | 39% | 40% | 6pp | Market hedges more than the model | Model closest |
| 4 July 2026 | Paraguay v France | Round of 16 | 0–1 Away win | 69% | 81% | 83% | 12pp | Market more confident than the model | Polymarket closest |
| 4 July 2026 | Canada v Morocco | Round of 16 | 0–3 Away win | 49% | 53% | 54% | 5pp | Market more confident than the model | Sources agreed |
| 3 July 2026 | Argentina v Cabo Verde | Round of 32 | 3–2 Draw | 10% | 11% | 11% | 6pp | Market hedges more than the model | All missed |
| 3 July 2026 | Colombia v Ghana | Round of 32 | 1–0 Home win | 86% | 67% | 68% | 19pp | Market hedges more than the model | Model closest |
| 3 July 2026 | Australia v Egypt | Round of 32 | 1–1 Draw | 29% | 33% | 33% | 27pp | Model and market disagree on the favourite | All missed |
| 2 July 2026 | Portugal v Croatia | Round of 32 | 2–1 Home win | 51% | 55% | 55% | 5pp | Market more confident than the model | Sources agreed |
| 2 July 2026 | Spain v Austria | Round of 32 | 3–0 Home win | 80% | 72% | 74% | 8pp | Market hedges more than the model | Model closest |
| 2 July 2026 | Switzerland v Algeria | Round of 32 | 2–0 Home win | 41% | 47% | 47% | 7pp | Market more confident than the model | Sources agreed |
| 1 July 2026 | England v DR Congo | Round of 32 | 2–1 Home win | 78% | 75% | 76% | 4pp | Market hedges more than the model | Sources agreed |
| 1 July 2026 | USA v Bosnia and Herzegovina | Round of 32 | 2–0 Home win | 66% | 69% | 70% | 5pp | Market more confident than the model | Sources agreed |
| 1 July 2026 | Belgium v Senegal | Round of 32 | 3–2 Draw | 32% | 29% | 29% | 7pp | Market more confident than the model | All missed |
| 30 June 2026 | France v Sweden | Round of 32 | 3–0 Home win | 80% | 74% | 77% | 7pp | Market hedges more than the model | Model closest |
| 30 June 2026 | Côte d'Ivoire v Norway | Round of 32 | 1–2 Away win | 52% | 45% | 45% | 8pp | Market hedges more than the model | Model closest |
| 30 June 2026 | Mexico v Ecuador | Round of 32 | 2–0 Home win | 36% | 42% | 43% | 8pp | Market more confident than the model | Sources agreed |
| 29 June 2026 | Germany v Paraguay | Round of 32 | 1–1 Draw | 28% | 19% | 19% | 13pp | Market more confident than the model | All missed |
| 29 June 2026 | Netherlands v Morocco | Round of 32 | 1–1 Draw | 32% | 30% | 30% | 11pp | Model and market disagree on the favourite | All missed |
| 29 June 2026 | Brazil v Japan | Round of 32 | 2–1 Home win | 39% | 57% | 57% | 18pp | Market more confident than the model | Sources agreed |
| 28 June 2026 | South Africa v Canada | Round of 32 | 0–1 Away win | 65% | 58% | 59% | 7pp | Market hedges more than the model | Model closest |
| 27 June 2026 | Algeria v Austria | Group J | 3–3 Draw | 32% | 43% | 43% | 15pp | Model and market disagree on the favourite | Markets closest |
| 27 June 2026 | Jordan v Argentina | Group J | 1–3 Away win | 80% | 81% | 84% | 4pp | Market more confident than the model | Polymarket closest |
| 27 June 2026 | Colombia v Portugal | Group K | 0–0 Draw | 32% | 26% | 25% | 16pp | Market more confident than the model | All missed |
| 27 June 2026 | DR Congo v Uzbekistan | Group K | 3–1 Home win | 24% | 56% | 56% | 32pp | Model and market disagree on the favourite | Markets closest |
| 27 June 2026 | Panama v England | Group L | 0–2 Away win | 63% | 82% | 85% | 19pp | Market more confident than the model | Polymarket closest |
| 27 June 2026 | Croatia v Ghana | Group L | 2–1 Home win | 78% | 54% | 54% | 25pp | Market hedges more than the model | Model closest |
| 26 June 2026 | Egypt v Iran | Group G | 1–1 Draw | 30% | 36% | 36% | 23pp | Model and market disagree on the favourite | All missed |
| 26 June 2026 | New Zealand v Belgium | Group G | 1–5 Away win | 65% | 82% | 82% | 18pp | Market more confident than the model | Sources agreed |
| 26 June 2026 | Cabo Verde v Saudi Arabia | Group H | 0–0 Draw | 31% | 28% | 28% | 10pp | Model and market disagree on the favourite | All missed |
| 26 June 2026 | Uruguay v Spain | Group H | 0–1 Away win | 71% | 64% | 65% | 7pp | Market hedges more than the model | Model closest |
| 26 June 2026 | Norway v France | Group I | 1–4 Away win | 55% | 58% | 60% | 6pp | Market more confident than the model | Sources agreed |
| 26 June 2026 | Senegal v Iraq | Group I | 5–0 Home win | 61% | 78% | 78% | 17pp | Market more confident than the model | Sources agreed |
| 25 June 2026 | USA v Türkiye | Group D | 2–2 Draw | 31% | 24% | 23% | 19pp | Model and market disagree on the favourite | All missed |
| 25 June 2026 | Paraguay v Australia | Group D | 0–0 Draw | 31% | 43% | 43% | 19pp | Model and market disagree on the favourite | Markets closest |
| 25 June 2026 | Curaçao v Côte d'Ivoire | Group E | 0–2 Away win | 64% | 83% | 83% | 20pp | Market more confident than the model | Sources agreed |
| 25 June 2026 | Ecuador v Germany | Group E | 2–1 Home win | 33% | 25% | 25% | 16pp | Market more confident than the model | All missed |
| 25 June 2026 | Japan v Sweden | Group F | 1–1 Draw | 24% | 28% | 27% | 19pp | Market hedges more than the model | All missed |
| 25 June 2026 | Tunisia v Netherlands | Group F | 1–3 Away win | 70% | 85% | 87% | 15pp | Market more confident than the model | Polymarket closest |
| 24 June 2026 | Mexico v Czechia | Group A | 3–0 Home win | 71% | 50% | 51% | 21pp | Market hedges more than the model | Model closest |
| 24 June 2026 | South Africa v South Korea | Group A | 1–0 Home win | 13% | 17% | 17% | 4pp | Market hedges more than the model | All missed |
| 24 June 2026 | Canada v Switzerland | Group B | 1–2 Away win | 28% | 40% | 40% | 12pp | Model and market disagree on the favourite | Markets closest |
| 24 June 2026 | Bosnia and Herzegovina v Qatar | Group B | 3–1 Home win | 47% | 67% | 68% | 21pp | Market more confident than the model | Sources agreed |
| 24 June 2026 | Scotland v Brazil | Group C | 0–3 Away win | 63% | 69% | 71% | 6pp | Market more confident than the model | Sources agreed |
| 24 June 2026 | Morocco v Haiti | Group C | 4–2 Home win | 74% | 81% | 83% | 8pp | Market more confident than the model | Polymarket closest |
| 23 June 2026 | Portugal v Uzbekistan | Group K | 5–0 Home win | 66% | 80% | 82% | 14pp | Market more confident than the model | Sources agreed |
| 23 June 2026 | Colombia v DR Congo | Group K | 1–0 Home win | 75% | 63% | 65% | 12pp | Market hedges more than the model | Model closest |
| 23 June 2026 | England v Ghana | Group L | 0–0 Draw | 13% | 15% | 14% | 9pp | Market hedges more than the model | All missed |
| 23 June 2026 | Panama v Croatia | Group L | 0–1 Away win | 46% | 62% | 63% | 16pp | Market more confident than the model | Sources agreed |
| 22 June 2026 | France v Iraq | Group I | 3–0 Home win | 83% | 88% | 90% | 7pp | Market more confident than the model | Sources agreed |
| 22 June 2026 | Norway v Senegal | Group I | 3–2 Home win | 43% | 42% | 42% | 5pp | Market hedges more than the model | Sources agreed |
| 22 June 2026 | Argentina v Austria | Group J | 2–0 Home win | 76% | 61% | 64% | 15pp | Market hedges more than the model | Model closest |
| 22 June 2026 | Jordan v Algeria | Group J | 1–2 Away win | 45% | 61% | 61% | 16pp | Market more confident than the model | Sources agreed |
| 21 June 2026 | Belgium v Iran | Group G | 0–0 Draw | 32% | 21% | 21% | 28pp | Market more confident than the model | All missed |
| 21 June 2026 | New Zealand v Egypt | Group G | 1–3 Away win | 47% | 60% | 60% | 13pp | Market more confident than the model | Sources agreed |
| 21 June 2026 | Spain v Saudi Arabia | Group H | 4–0 Home win | 90% | 87% | 88% | 4pp | Market hedges more than the model | Sources agreed |
| 21 June 2026 | Uruguay v Cabo Verde | Group H | 2–2 Draw | 21% | 22% | 22% | 8pp | Market hedges more than the model | All missed |
| 20 June 2026 | Germany v Côte d'Ivoire | Group E | 2–1 Home win | 60% | 63% | 64% | 6pp | Market more confident than the model | Sources agreed |
| 20 June 2026 | Ecuador v Curaçao | Group E | 0–0 Draw | 16% | 10% | 9% | 7pp | Market more confident than the model | All missed |
| 20 June 2026 | Netherlands v Sweden | Group F | 5–1 Home win | 67% | 55% | 55% | 13pp | Market hedges more than the model | Model closest |
| 20 June 2026 | Tunisia v Japan | Group F | 0–4 Away win | 71% | 62% | 63% | 8pp | Market hedges more than the model | Model closest |
| 19 June 2026 | Scotland v Morocco | Group C | 0–1 Away win | 61% | 55% | 56% | 5pp | Market hedges more than the model | Model closest |
| 19 June 2026 | Brazil v Haiti | Group C | 3–0 Home win | 76% | 86% | 88% | 11pp | Market more confident than the model | Sources agreed |
| 19 June 2026 | USA v Australia | Group D | 2–0 Home win | 34% | 60% | 61% | 26pp | Model and market disagree on the favourite | Markets closest |
| 19 June 2026 | Türkiye v Paraguay | Group D | 0–1 Away win | 20% | 24% | 25% | 4pp | Market hedges more than the model | All missed |
| 18 June 2026 | Czechia v South Africa | Group A | 1–1 Draw | 29% | 26% | 25% | 3pp | Market more confident than the model | All missed |
| 18 June 2026 | Mexico v South Korea | Group A | 1–0 Home win | 60% | 47% | 48% | 13pp | Market hedges more than the model | Model closest |
| 18 June 2026 | Switzerland v Bosnia and Herzegovina | Group B | 4–1 Home win | 76% | 62% | 62% | 15pp | Market hedges more than the model | Model closest |
| 18 June 2026 | Canada v Qatar | Group B | 6–0 Home win | 86% | 75% | 75% | 11pp | Market hedges more than the model | Model closest |
| 17 June 2026 | Portugal v DR Congo | Group K | 1–1 Draw | 20% | 17% | 16% | 5pp | Market hedges more than the model | All missed |
| 17 June 2026 | Uzbekistan v Colombia | Group K | 1–3 Away win | 61% | 69% | 71% | 8pp | Market more confident than the model | Sources agreed |
| 17 June 2026 | England v Croatia | Group L | 4–2 Home win | 53% | 56% | 56% | 4pp | Market more confident than the model | Sources agreed |
| 17 June 2026 | Ghana v Panama | Group L | 1–0 Home win | 11% | 44% | 44% | 38pp | Model and market disagree on the favourite | Markets closest |
| 16 June 2026 | France v Senegal | Group I | 3–1 Home win | 65% | 65% | 66% | 3pp | Market more confident than the model | Sources agreed |
| 16 June 2026 | Iraq v Norway | Group I | 1–4 Away win | 65% | 80% | 82% | 15pp | Market more confident than the model | Sources agreed |
| 16 June 2026 | Argentina v Algeria | Group J | 3–0 Home win | 74% | 69% | 70% | 6pp | Market hedges more than the model | Model closest |
| 16 June 2026 | Austria v Jordan | Group J | 3–1 Home win | 45% | 71% | 73% | 27pp | Market more confident than the model | Sources agreed |
| 15 June 2026 | Belgium v Egypt | Group G | 1–1 Draw | 29% | 24% | 24% | 5pp | Market more confident than the model | All missed |
| 15 June 2026 | Iran v New Zealand | Group G | 2–2 Draw | 25% | 28% | 28% | 13pp | Market hedges more than the model | All missed |
| 15 June 2026 | Spain v Cabo Verde | Group H | 0–0 Draw | 9% | 8% | 7% | 4pp | Market hedges more than the model | All missed |
| 15 June 2026 | Saudi Arabia v Uruguay | Group H | 1–1 Draw | 24% | 22% | 21% | 3pp | Market more confident than the model | All missed |
| 14 June 2026 | Germany v Curaçao | Group E | 7–1 Home win | 84% | 92% | 94% | 10pp | Market more confident than the model | Sources agreed |
| 14 June 2026 | Côte d'Ivoire v Ecuador | Group E | 1–0 Home win | 17% | 26% | 26% | 16pp | Market hedges more than the model | All missed |
| 14 June 2026 | Netherlands v Japan | Group F | 2–2 Draw | 32% | 27% | 27% | 14pp | Model and market disagree on the favourite | All missed |
| 14 June 2026 | Sweden v Tunisia | Group F | 5–1 Home win | 41% | 50% | 51% | 9pp | Market more confident than the model | Sources agreed |
| 13 June 2026 | Qatar v Switzerland | Group B | 1–1 Draw | 16% | 15% | 14% | 3pp | Market hedges more than the model | All missed |
| 13 June 2026 | Brazil v Morocco | Group C | 1–1 Draw | 32% | 25% | 24% | 20pp | Market more confident than the model | All missed |
| 13 June 2026 | Haiti v Scotland | Group C | 0–1 Away win | 45% | 61% | 62% | 16pp | Market more confident than the model | Sources agreed |
| 13 June 2026 | Australia v Türkiye | Group D | 2–0 Home win | 32% | 19% | 17% | 20pp | Market more confident than the model | All missed |
| 12 June 2026 | Canada v Bosnia and Herzegovina | Group B | 1–1 Draw | 18% | 27% | 27% | 28pp | Market hedges more than the model | All missed |
| 12 June 2026 | USA v Paraguay | Group D | 4–1 Home win | 44% | 49% | 50% | 5pp | Market more confident than the model | Sources agreed |
| 11 June 2026 | Mexico v South Africa | Group A | 2–0 Home win | 83% | 68% | 69% | 15pp | Market hedges more than the model | Model closest |
| 11 June 2026 | South Korea v Czechia | Group A | 2–1 Home win | 47% | 36% | 36% | 11pp | Market hedges more than the model | Model closest |
Probabilities are each source's frozen pre-match figure on the outcome that actually occurred — read from the ledger, not recomputed after the fact. "Gap" is the largest single-outcome difference between the model and the sportsbook (the same metric the divergence flag uses; ≥ 15pp is flagged). A dash means no market was posted for that match, so it can't be scored. Download the full archive as CSV.