Odds and the recommended pick

A probability on its own isn't a recommendation. The last step is comparing the model's view to the free odds and naming a pick.

The free odds

For each fixture the model fetches free pre-match 1X2 and Over/Under 2.5 odds from API-Football, taking the best (highest) price across a handful of bookmakers per outcome and keeping the per-book breakdown so the card can show where the best line sits.

How the pick is chosen

The recommended pick is deliberately simple and transparent:

• 1X2 — the side (home, draw or away) with the highest model probability, shown with its odds.
• Over/Under — Over if the predicted total goals is above 2.5, otherwise Under, shown with the matching price.

So the pick is the model's most-likely outcome, paired with what the market will pay for it. The odds aren't used to change the prediction — they're shown next to it so you can judge whether the price is worth the confidence (the difference between a likely outcome and a good bet is the heart of the football-betting track).

A frozen snapshot

When a pick is made, the model also stores an immutable pre-match snapshot — the probabilities, the recommended side, the odds at the time, and the Over/Under read. That snapshot is what later gets settled against the real result to build the win/loss and ROI record, so the track record reflects what was actually called before kickoff.

The pick is the most-likely side at the price on offer. Whether that price is generous is a separate question — and one the model leaves to you.