2026 FIFA World Cup Predictions API
Build World Cup 2026 win-probability models with historical match data, xG, team strength ratings, and live odds. Everything you need to feed a Poisson, Elo, or Dixon-Coles model, then validate against closing line value from Pinnacle.
From data to model
Pull 10 years of historical match results, xG, and lineups, then train your model offline. The API gives you the inputs in clean JSON — you bring the math. Our free /tools/poisson-score-predictor is a worked client-side example.
Validate with closing line value
Compare your predicted probabilities to Pinnacle's closing prices via the odds endpoint. CLV is the most reliable signal that a prediction model is genuinely sharp.
Common modelling approaches
Poisson models predict goal counts from team attack/defence rates. Dixon-Coles corrects Poisson's low-score independence assumption. Elo-style ratings adapt team strength over time. Bivariate Poisson and goal-time models add temporal structure. All of these are computable from the data this API returns.
Probability outputs to surface in your UI
Typical model outputs to display: P(home win) / P(draw) / P(away win), expected score distribution (0-0 through 5-5+), most-likely scoreline, BTTS probability, over/under 2.5 probability, and tournament-progression chances. All are derivable from the same goal-expectancy parameters.
/football/matches
Use competition_id={COMPETITION_ID} to scope this endpoint to the 2026 FIFA World Cup. The same endpoint works for 1,195 other competitions.
// Historical xG and result data for predictive modelling
const response = await fetch(
'https://api.thestatsapi.com/api/football/matches?competition_id={COMPETITION_ID}&season=2022&include=xg',
{ headers: { 'Authorization': 'Bearer YOUR_API_KEY' } }
);
const { data } = await response.json();
// Feed historical xG + results into Poisson, Elo, or Dixon-ColesWhat the World Cup 2026 predictions API returns
Selected fields from the JSON response. Full schema and example payloads are in the developer docs.
| Field | Type | Description |
|---|---|---|
| season | integer | Year (e.g. 2022, 2018) for historical World Cups. |
| match.xg | object | { home, away } — total expected goals per team. |
| match.shots | array | Per-shot data: minute, player, xg, on_target, body_part. |
| team.elo | decimal | Rolling Elo rating at match time. |
| team.rest_days | integer | Days since last fixture — useful for fatigue modelling. |
| odds.closing | object | Pinnacle closing prices for CLV benchmark. |
| venue.altitude_m | integer | Altitude in metres — relevant for Mexico City fixtures. |
What developers build with the World Cup 2026 predictions API
Bracket-prediction Monte Carlo
Simulate the full tournament 10k+ times to estimate each team's win probability.
Win-probability widgets
Render pre-match P(home/draw/away) updated as lineups land.
Edge-finding bots
Compare model P() to bookmaker implied probability and surface positive-EV bets.
Editorial tipster posts
Generate auto-published preview articles with model-derived predictions and confidence intervals.
Implementation notes
- Always benchmark new models against Pinnacle closing prices — it is the cleanest no-vig signal.
- Use rolling-window xG (last 10–20 matches) rather than full-season xG to capture form.
- Down-weight friendly results in your training set — friendlies are notoriously noisy.
- For knockout matches, model 90 minutes + extra time + penalties separately and combine probabilistically.
- Publish methodology and feature importance for your model — it builds trust and is a strong SEO play.
Related World Cup data
Poisson Score Predictor
Free client-side tool: input xG, get score distribution.
Odds Converter
Convert odds and implied probability.
Odds API
Live and pre-match odds from 4 bookmakers.
2022 history
Train on the previous tournament.
Closing line value
CLV workflows for model validation.
Top scorers
Player-level xG for Golden Boot prediction.
Frequently Asked Questions
What model should I use for World Cup predictions?
Poisson and Dixon-Coles models work well for low-scoring sports like football. Elo ratings are simpler and adapt over time. Combining xG-based goal expectancy with Elo team strength is a common starting point.
Does the API provide pre-built predictions?
No — the API provides the raw inputs (results, xG, odds, team stats). You build the model. This gives you full control over assumptions and lets you ship models other people cannot replicate from public predictions.
How much historical data is available?
10 years of match data including xG for major competitions, and full World Cup historical records back to 1930.
Can I compare my predictions to bookmaker odds?
Yes. Use the odds endpoint to get Pinnacle, Bet365, Betfair, and Kambi prices. Pinnacle's closing line is the standard sharp benchmark for model validation.
How do I model the 48-team / Round of 32 format?
Estimate per-match goal expectancies, then run Monte Carlo simulations of the entire bracket from group stage onward. Repeat 10k+ times to derive each team's probability of advancing at each stage and ultimately winning.
What features matter most for football models?
In order of typical predictive power: recent xG for/against, Elo or SPI-style team strength, opponent quality adjustment, home/neutral advantage (often weak at World Cups), rest days, and squad availability.
Where can I see a worked end-to-end example?
See our blog post on building a World Cup ML prediction model and the /tools/poisson-score-predictor tool for a runnable client-side version.
Start building with the World Cup 2026 API
7-day free trial. All endpoints included on every plan.