We publish our numbers. KickoffAPI's match projections are backtested leak-free — every prediction uses only matches played before kickoff, exactly as it runs live — and scored against actual results. Here's how the model does.
Scored with the Ranked Probability Score (RPS) — the standard metric for 1X2 forecasts, where lower is better and 0 is perfect. Strong public models land around 0.19–0.21; bookmakers (the strongest forecasters, with a margin baked in) around 0.19. We beat a naive base-rate baseline in every league.
| League | KickoffAPI RPS | Baseline RPS | Top-pick accuracy |
|---|---|---|---|
| Premier League | 0.207 | 0.234 | 53.5% |
| La Liga | 0.197 | 0.223 | 50.0% |
| Serie A | 0.200 | 0.225 | 52.0% |
| Bundesliga | 0.200 | 0.229 | 51.7% |
| Ligue 1 | 0.226 | 0.237 | 45.3% |
A Dixon-Coles model — team attack and defence strengths with exponential time-weighting so recent form counts more, a low-score correlation correction that gets draws and tight games right, and shrinkage toward the league average so thin samples and one-team-dominant leagues don't skew it. It's our own model, computed from our own results data. Every new version is validated on held-out history and only ships if it beats the one before — so accuracy only moves one way.
Available through the API: /api/v2/fixtures/:id/projection (1X2, expected goals, BTTS, over/under) and /api/v2/insights/season-sim (title, top-4 and relegation odds). See the v2 reference.
On a leak-free backtest of the 2023 big-5 leagues, the model scores RPS around 0.20 and picks the correct result ~50–53% of the time, beating a naive baseline in every league.
No. These are model-derived statistical projections for analysis and product use — not betting advice, and not a guarantee of any outcome.
Through the KickoffAPI v2 API on any paid plan. Start free and read the v2 reference.
Transparent, backtested football predictions through one fast API. Free tier, no credit card.