About
Sports data, models, and AI should belong to everyone
Professional clubs and national federations have spent the last decade buying their way into analytics: proprietary tracking feeds, internal model stacks, dedicated data-science staff, and tools the public never sees. OnThePitch opens that up: the data, the models, and the AI behind them, starting with the 2026 FIFA World Cup.
OnThePitch
What we publish
A model for the World Cup that answers the questions fans actually ask: who wins, who qualifies from the group, who starts, and who scores? Win probabilities for every fixture, predicted starting XIs for every nation, and the full methodology behind it all. Built from public data, refreshed daily, documented openly.
See /docs/methodology/ for the full write-up, /posts/ for short-form research notes, and /data/ for the export endpoints.
How to read a probability
Every number on the site is a probability: a long-run frequency, not a forecast of what will definitely happen. When the model gives a team an 18% chance of winning, it means that across many comparable matches an outcome like this lands about once in every five. Low-probability results happen all the time; that's exactly what an 18% means.
Because the numbers are frequencies, we check them the same way: by comparing what the model said against what actually happened, match after match. The methodology walks through how the probabilities are built and calibrated, and every term or column you'll meet along the way has its own definition.
Who we're for
Fantasy players who want to know who'll actually start. Football journalists looking for a citable number instead of a vibes-based take. Developers who want clean per-fixture data via API. And anyone who watches the game and wants to see what the numbers say before kickoff.
Why we're doing this
A modern Premier League club runs five-figure-per-month data contracts, employs full-time analysts, and ships tactical reports their fans will never read. Most of what they produce isn't secret, but it's expensive to assemble, and the assembly is what costs.
The argument we'd make is that the assembly should be a public good. If anyone with a browser can read a calibrated tournament forecast, the conversation about a fixture stops being what does the bloke on TV think and starts being here's the model's view, here's where it's confident, here's where it isn't. That's a better conversation, and it's the one this project is trying to enable.
How we work
Open sources
Everything the model reads is publicly available: FIFA fixtures, Wikipedia squad pages, FBref via the worldfootballR ecosystem, public club-season stats. The full source list is in the methodology page, and any reader can reproduce the inputs from the same public archives.
Methodology in the open
Every probability on the site comes from a model whose architecture, features, training procedure, and limitations are written up at /docs/. If a number on the page surprises you, the methodology should answer where it came from.
Free baseline access
The headline forecast (tournament winner, group standings, the knockout path) is free for every reader, along with predicted line-ups, squads, player ratings, and the full methodology. Deeper match-by-match breakdowns sit behind a one-time Pass that funds the work.
Editorial scope
OnThePitch is a football research publication in the same tradition as FiveThirtyEight or FBref. The product is the model's probabilities, the methodology behind them, and the per-team and per-player breakdowns. Revenue funds the research.
Get in touch
Found a bug, want to suggest a feature, or have a research question? The fastest path is the feedback form.
Who's behind it
OnThePitch is an independent sports analytics project.
The model has been scored on 987 out-of-sample matches across 24 major tournaments since 2014. Every probability published during this World Cup is graded against the real result.
See the full track record on the accuracy page.