YR Sports Lab
What Scorelines Will the 72 World Cup 2026
Group Games Produce? Match-by-Match Predictions and a Score-Guessing Game
The title odds are done, but what you really want to know is: by how many goals does Argentina beat Algeria? And what about France against Senegal in the group of death? This article reuses the cross-validated double-Poisson scoreline engine from our title-odds article, converting the composite strength gap between any two of the 48 teams into an expected-goals figure for each side, then expanding that into a full scoreline probability grid. The output: the most likely score, the win-draw-loss split and both teams' scoring probabilities for every one of the 72 group games across 12 groups. At the end there is a score-guessing game, so you can see whether you can beat the model.
Match-by-match scorelines: six games per group, 72 in total
Use the group tabs to switch between pools. On each card the large figure is that game's most likely scoreline (the model mode), the bar underneath is the win / draw / loss split (blue = home win, grey = draw, amber = away win; "home" and "away" here are only the fixture-list designations, since most World Cup group games are played at neutral venues), and below that are the three highest-probability scorelines. A reminder: the most likely scoreline usually carries only a 12–19% chance. Football is a low-scoring, highly random sport, so "most likely" is a long way from "likely".
Most common scorelines: 1-0 is the World Cup's default setting
Adding up the scoreline probabilities across all 72 games gives us the expected share of matches that finish on any given scoreline. The result fits football's low-scoring nature perfectly: 1-0 tops the list at 18.7%, with 2-0 and 2-1 close behind, and every one of the top six involves three goals or fewer. Blowouts (4-0, 5-1 and the like) do turn up a few times at every tournament, but they are spread across so many possible scorelines that no single one carries much probability. Bar length shows each scoreline's expected share of the 72 games.
How many goals in a game? The total-goals distribution
Summing both teams' expected goals in each game gives a probability distribution over total goals. Two-goal games are the most common, followed by three and then one, and "over 2.5 goals" (three or more) accounts for roughly half of all games. Goalless 0-0 grinds come in around 7%, while games with four goals or more add up to a little over twenty percent. The bars show each total's expected share of the 72 games.
Why does the average land exactly on 2.65 goals?
That is no coincidence, it is a calibration choice in the model: the double-Poisson fixes the sum of the two teams' expected goals at 2.65 (the long-run average goals per game across recent World Cups), then splits those 2.65 goals between the stronger and weaker side according to the strength gap. So what the model predicts is who scores and how the scoreline divides, not whether a given game will be a shootout. That is a deliberate trade-off, explained in the methodology below.
Marquee matches: scoreline scripts for four headline group games
Here are the four most talked-about fixtures with their full top-six scorelines laid out. Notice that the more evenly matched the tie (Group I, the group of death, is the clearest case), the lower the probability of the most likely scoreline and the more the probability mass spreads out. That is the mathematical footnote to the old truth that big games are the hardest to call.
Score-guessing challenge: can you beat the model?
The game draws 8 fixtures at random from the 72 group games. Use the + and − buttons to set a scoreline, then submit. Scoring: +5 points for an exact scoreline, +2 points for calling the right result only, and nothing if you miss both. Every submission reveals the probability the model assigns to your scoreline, along with the score it favours itself. Put your football instincts up against cold, hard Poisson.
Scoreline probabilities are computed live from each game's double-Poisson λ values, from the same source as the predictions above. Fixtures are drawn with an unseeded shuffle, so every round is different.
Methodology and limitations: how each scoreline is derived
1. The double-Poisson scoreline model
Goals in football approximately follow a Poisson distribution, a classic result in sports statistics (Maher, 1982; Dixon & Coles, 1997). This article reuses the scoreline engine of Model 5 (M5, composite-strength Poisson) from our title-odds article: take the composite strength gap Δr between the two teams (the centred average of four models built on Elo, FIFA points, squad market value and historical record, already including the +80 host bonus), convert it into an expected-goals differential, then split that between the sides:
| Step | Formula | Notes |
|---|---|---|
| Goal differential μ | μ = clamp(Δr / 170, ±3) | Every 170 rating points is worth roughly one goal of expected margin, capped at ±3 |
| Home expectancy λA | λA = max(0.18, (2.65 + μ)/2) | The stronger side takes the larger share of the 2.65 goals |
| Away expectancy λB | λB = max(0.18, (2.65 − μ)/2) | The weaker side takes the smaller share, with a 0.18 floor |
| Scoreline probability | P(x:y) = Pois(x; λA) · Pois(y; λB) | The two teams' goals are treated as independent, expanded into an 11×11 probability grid |
Win-draw-loss, both teams to score, total goals, over 2.5 goals and every other figure here are all summed from the same scoreline grid, so they are internally consistent: each grid's probabilities sum to 1, and win + draw + loss is always exactly 100%. The most likely scoreline is simply the highest-probability cell in the grid.
2. Input data and validation
Strength ratings and group data were collected and cross-checked before kick-off on 2026-06-11: the 12 groups were verified as identical across two sources, Wikipedia and NBC Sports; Elo comes from eloratings.net, FIFA points from the official 11 June ranking, and market values from the total value of each 26-player squad on Transfermarkt. These inputs are exactly the same as in the title-odds article, which already validated the composite strength ratings against the betting market (Spearman 0.81) and five public expert models. Rather than repeat that work, this article simply recasts the same ratings in scoreline terms.
3. Limitations (read before the predictions)
4. Sources
- Maher, M. J. (1982). Modelling association football scores. Statistica Neerlandica, 36(3), 109–118.
- Dixon, M. J. & Coles, S. G. (1997). Modelling Association Football Scores and Inefficiencies in the Football Betting Market. Journal of the Royal Statistical Society: Series C, 46(2), 265–280: rss.onlinelibrary.wiley.com
- Strength ratings and validation sources from our companion article, "World Cup 2026 Title Odds": yrsportsblog.pages.dev/en/posts/worldcup-2026-prediction/
- World Cup 2026 group draw (Wikipedia × NBC Sports, cross-checked): en.wikipedia.org/wiki/2026_FIFA_World_Cup_draw
- World Football Elo Ratings: eloratings.net