121 possible scores

Football forecastingTwo modelsWorld Cup 2026 · Premier League 2026–27

Every match has 121 possible scores.

This project gives every one of them a probability, from 0–0 to 10–10, before kickoff. Then it publishes how that went.

Live sculptureλ 1.55 · 1.10
Home win47.0%
Draw26.8%
Away win26.2%

The engine

How a single result is calculated.

Six steps turn past results into a probability for every possible score. Both models follow them; each tunes the details for its own competition.

Step 1

Only what was known at the time

The model reads past results, but only those that had actually happened before the forecast was made. Anything later is invisible to it, so it can never peek at the future.

Step 2

Two strengths per team

From those results it learns an attack and a defence strength for every team, with an honest measure of uncertainty. Teams with little data are pulled gently towards average.

Step 3

Expected goals for this match

Your attack, their defence, a league-wide base rate and home advantage combine into an expected number of goals for each side.

log λ = μ + home + attack − defence

Step 4

All 121 scores

Goals follow Poisson curves, with the Dixon–Coles correction for how often low scores like 0–0 and 1–1 really happen. Every score from 0–0 to 10–10 gets a probability.

Step 5

Read off every market

Home win, draw and away win are sums of regions of the same grid, and so are totals and goal margins. They can never contradict each other.

Step 6

Play it all out thousands of times

To answer bigger questions, like who wins the tournament or the league, the model simulates every remaining match many thousands of times and counts the outcomes.

Try it

Build a match. Watch the 121 scores move.

Set each team's attack and defence and switch home advantage on or off. The engine turns them into expected goals, then into the full score landscape and the three outcomes.

Home team

Away team

Home advantage
λ home = e^(μ 0.20+home 0.18+att 0.20−def 0.00)1.73
λ away = e^(μ 0.20+att 0.00−def 0.10)1.11
Home win–
Draw–
Away win–

Most likely single score: –

Illustrative values chosen for teaching: a base rate μ of 0.20 and a home advantage of 0.18 on the log scale. The real models learn these from data.

Same engine, different jobs

What changes between the two models.

The core never changes. What each model learns from, how it treats home advantage, and how it is judged are tuned to its competition.

World Cup modelPremier League model
Learns from49,000+ international results4,560 Premier League matches, 2014/15 on
Elo anchorK = 40, scaled by match importanceK = 20, league matches only
Home advantageHosts get 1.4× the learned edge; elsewhere each side gets halfLearned from the data, applied to every home side
New or thin teamsUncertainty widened for teams with little dataPromoted clubs start 75 Elo below average; unseen clubs start from the prior
Simulation20,000 whole tournaments, extra time and penalties included20,000 whole seasons under league tiebreak rules
How often it re-fitsAbout 40 daily refits through the tournamentA new forecast after new results arrive
Judged againstTied plain Elo; the market was better by 0.010Slightly ahead of Elo, not proven; the market is ahead by 0.0065
StatusFinished, archived with a DOILive, scored match by match
NextThe World Cup model→