Belgium 4-1 Italy: result, our prediction
Pre-match model estimates · MODELS TOP 3
Based on the Road2Sushi model. Not guaranteed outcomes.
Prediction snapshot · Model generated .
Post-match analysis
Judged against the prediction this page published before kickoff. Right or wrong, it stays on the record.
A call is the side our published probability favoured. If we leaned towards under and three goals went in, that counts as a miss.
Match data
| BelgiumBelgium | Statistic | ItalyItaly |
|---|---|---|
| 4 | Goals | 1 |
The result, on the record
Recorded tally 1 of 3 published calls on this match were right.
Published calls · recorded grading
| # | Market | Our pick | Our chance | Recorded grade |
|---|---|---|---|---|
| 1 | Match result | Pick: Belgium | Chance: 51.4% | Correct |
| 2 | Total goals 3.5 | Pick: Under 3.5 | Chance: 77.8% | Wrong5 goals |
| 3 | Both teams to score | Pick: No | Chance: 53.8% | Wrong |
Our verdict
Road2Sushi's own call on this match: the likeliest result, how confident the model is, its strongest signals and the reasons behind them.
Every percentage is a model estimate and every fair price is 1 divided by that estimate. Neither is a bookmaker offer, and a model call is not a proven edge over the market.
Model pick
Top signals
Model read
Form & head-to-head
Recent form uses each club’s available results in their stored order. Select a W/D/L result for details.
Head-to-head figures use their own recorded sample. Where individual results are available, the figures and journal cover the same meetings. Missing results are never counted as draws.
W
L
W
L
W
3 wins · 0 draws · 2 losses
L
W
D
W
W
3 wins · 1 draw · 1 loss
Head-to-head
Last 5 meetings
Odds: model vs bookmaker
Every price for this match in one place: the model's probabilities next to the bookmaker's, the bookmaker's prices with our gap, and the model's own fair odds for each market.
Bookmaker probabilities are de-vigged (the margin is removed) and come from one bookmaker. A gap between the model and that bookmaker is a disagreement, not a proven edge. Fair odds are 1 divided by the model probability, not a bookmaker quote.
Model vs market
Pre-match estimates, recorded before the match.
1X2, Goals, Corners, Cards and Shots: Bet365 prices via API-Football, observed 15 Nov 2026, 15:32 UTC, with the bookmaker margin taken out.
A market probability is the bookmaker's own price with its margin taken out, so the outcomes of each line add up to 100%. A gap from our model is a disagreement, not a proven edge: when we tested 1X2 and goal totals on matches the model had never seen, the bookmaker's closing price was more accurate. For corners, cards and shots we have no such test.
Cards: the bookmaker settles on its own card count, which may include red cards. Our model counts yellow cards only.
Shots: the bookmaker settles on its own shot count, which can differ from the match statistics our model learns from.
Every market price shown was recorded before kick-off.
1X2
Goals
Goals
Both teams combined.
Read context
Both teams combined
BTTS · Yes · Strongest read · Close call
Method, limits & performance
How this works (honestly)
We are transparent about our method, our limits and how we measure performance.
Every senior international since 1990 from an open CC0 results archive, plus our saved provider results. No odds, no xG, no lineups: none of them exist for national teams in our sources.
One Elo rating across all national teams, mapped to match odds with a Poisson model. Only markets that passed a preregistered out-of-sample test are shown: the match result, over 3.5 goals and both teams to score. Over 1.5 and over 2.5 goals failed their calibration test and are not published.
There is no bookmaker price in this model and no betting call. These matches are outside our AI experts, coupons and betting record.
Tested out of sample on 7,576 internationals: log loss 0.865 against 1.044 for a no-information baseline, calibration error 0.009. Ratings as of 2026-11-15.
Not financial advice. No model beats the sharp closing market on public data. The value here is honest probabilities, No-bet calls, and CLV history.
