How Footyntel's projections work

Footyntel turns public match data into probabilities — the chance each club wins its league, finishes in the qualification places, or goes down, plus a win/draw/win forecast for every remaining match. Everything here is a transparent statistical model. It is not betting advice and it never invents results.

1. The data

We use football-data.org for standings and fixtures across 9 round-robin leagues. Scores are delayed, not live — we are a projection and analysis site, not a livescore service. Numbers refresh roughly hourly.

2. Team ratings (attack & defence)

Each club gets an attack rating (how many goals it tends to score vs the league average) and a defence rating (how many it concedes). We don't trust a handful of early games on their own, so each rating is a blend:

• This season's results — goals scored and conceded per game so far.
• Last season's final record — the same club's attack/defence from the previous campaign (newly-promoted sides get a modest "promoted team" prior instead).

The blend weight is games / (games + 8): early in the season the prior does most of the work, and by spring it's almost entirely this season's form. This is the same Bayesian idea public models like the old FiveThirtyEight used, and it stops early-season flukes producing silly numbers.

3. Match forecasts

From the two teams' ratings and a home-advantage factor we get an expected goals value for each side, then use a Poisson model over every plausible scoreline to get win / draw / win probabilities, the most likely score, over/under 2.5 goals and both-teams-to-score.

4. Season simulation

To project the table we play out every remaining fixture 50,000 times. In each simulated season we also re-draw each team's strength from a range (wider when fewer games have been played) so the uncertainty is carried through honestly. The share of simulations in which a club finishes 1st, in the qualification places, or in the bottom few becomes its title / qualification / relegation percentage.

5. Tiebreakers

When simulated teams finish level we order them by points, then goal difference, then goals scored. Some leagues (La Liga, Serie A) officially use head-to-head first — we simplify to goal difference, which barely changes the aggregate probabilities.

6. What this is — and isn't

• It is a transparent, repeatable estimate of how a season could unfold.
• It isn't a certainty, a tip, or betting advice. Early-season numbers are deliberately wide.
• We never fabricate scores. Results come from the data; projections are clearly labelled as a model.

7. Accuracy tracking (coming soon)

We log a snapshot of every projection each day. Once there's a season's worth of history we'll publish how well-calibrated the model has been — scored against real outcomes — right here. No other free football model shows its own report card; we will.

8. Open data

Every league's projection is free to download as JSON and CSV (CC BY 4.0). See the "Open data" links on any league page.