Data, models,
and the limits of both.

Five seasons of historical results — roughly 16,000 matches across eight European leagues — sourced from Football-Data.co.uk. That covers goals, cards, corners, referee assignments, and closing odds from every major book.

Expected-goals (xG) data comes from Understat for the Premier League, La Liga, Serie A, Bundesliga and Ligue 1. Shot-level breakdowns — open play, set pieces, corners, free kicks, penalties — feed directly into the prediction layer.

Live prices come from The Odds API, preferring Pinnacle as the sharp baseline. Odds refresh on every pipeline run — currently daily, fixture-aware so we only call the API for leagues that actually have matches in the next 72 hours.

Predictions come from a probability ensemble — multiple models working together, each trained on five seasons of historical data. The ensemble combines team strength, recent form, expected goals, shot quality, and other signals into a single set of calibrated probabilities.

Outputs are calibrated to correct for known biases in football prediction — the common draw-underconfidence problem, for example — and blend weights are tuned per league to reflect each competition's characteristics. The Bundesliga and the Championship behave very differently from each other, and the model knows that.

We don't publish the exact model composition — that's our secret sauce. But we do publish every prediction we make, and you can judge the model on its results, not its marketing.

The ensemble has been tested on 2,171 completed matches using expanding windows — meaning the model never gets to peek at the future. At each point in time it only sees what would have been available on that date.

Backtest results are a guide, not a guarantee. Past performance does not predict future performance — especially over small samples.

Finding Space is built by a one-person team. The same person writes the data pipeline, trains the models, designs the app, and answers the support emails. That's a deliberate choice — it keeps the product opinionated, the roadmap honest, and the cost low enough that we don't need to chase volume with hype.

It's part of a small studio building data-driven sports tools across football, NFL and golf. All three projects share the same principle: probability over punditry, calibration over confidence, transparency over salesmanship.

This is research, not advice. Every prediction carries uncertainty. Only stake what you can genuinely afford to lose, and never chase losses.

If gambling is affecting you or someone close to you, free and confidential support is available at BeGambleAware.org.

18+ only. Gambling can be addictive — please play responsibly.