A statistical model rates every team from thousands of real results, prices a dozen markets as calibrated probabilities, and has AI explain each one in plain language — grounded strictly in the data. Then every call is graded in public.
Every number on ModelXI traces back to real match data. Here's the whole pipeline, start to finish.
A Dixon-Coles statistical model rates every team's attack and defence from thousands of real results across the last two seasons.
From that goal model it computes calibrated probabilities for the result, handicaps, totals, both-teams-to-score, correct scores and more.
A language model turns the numbers into a short, readable preview — grounded strictly in the model's output and real form.
Every match gets a headline call, full win probabilities, and a dozen graded markets underneath.
Illustrative examples of the format. Live calls update every day on the dashboard.
We publish every result instead. This is the honest answer to "how good is it?"
A prediction is only worth anything if its confidence is real. The test isn't whether we sound confident — it's whether a "70%" call actually happens about 70% of the time, over many matches.
That's why every single call the model makes is graded against the real final score and shown publicly. No cherry-picking, no hidden misses. You judge us on the record.
See the live track recordWe show you probabilities and the reasoning behind them. The decision is always yours.
Every plan runs on the same honest model. Paid tiers just remove the limits and add depth.
Prices in EUR. Paid plans open soon — the full dashboard is free while ModelXI is in early access.
No. ModelXI is statistical analysis for information — probability estimates and reasoning. It doesn't tell you to place a bet and it doesn't guarantee outcomes. What you do with it is your call.
No, and anyone who does is lying. Football is uncertain by nature. We estimate probabilities as honestly as we can and show our reasoning; we don't sell certainty.
You watch the track record. Every call the model makes is graded against the real final score and shown publicly, including calibration — whether a 70% call really happens about 70% of the time. It builds as matches are played, so you can verify it yourself.
Eighteen at launch — the big European leagues plus Brazil, Argentina, MLS, Liga MX, Liga I and others. Each has enough history for the model to be credible. More as it proves out.
Because our model doesn't measure them — and we won't fake numbers we can't stand behind. Those need different data and different models. They're on the roadmap, honestly labelled as coming.
Yes, for now. If premium tiers arrive later, the core analysis stays honest either way.
Eighteen leagues, a dozen markets each, refreshed daily — and every call graded in public.
Open the dashboard