From data to a prediction
What data goes in
4 min
A prediction is only as good as what feeds it. FinalSkore watches a game the way an attentive analyst would — through a steady stream of numbers, not a single snapshot.
The live inputs
During a game, the product ingests the live score and the game state: which quarter or half it is, how much time is left, and how big the gap is. It also tracks pace — roughly how fast both teams are using possessions — because the same lead means very different things in a fast game versus a slow one.
The market inputs
Alongside the game itself, FinalSkore pulls bookmaker odds, updated continuously as the game develops. Odds are useful because they encode what the wider betting market believes will happen. For games that haven't tipped off yet, it ingests scheduled fixtures with pre-game odds, across roughly 29 international leagues.
Why the mix matters
No single number tells the story. A 10-point lead, two minutes left, a slow pace, and a market that still favors the trailing team is a very different situation from the same lead early in a fast-paced first quarter. Combining score, time, pace, and odds is what lets a model reason about a game rather than just react to the last basket.
Garbage in, garbage out: the model reflects the data it sees, and live data is messy. Treat its output as an informed read, not a fact.