About iPick

What you will find here

• Match Center: the day's games, who the model picks in each one and how likely that pick is to win.
• Predictions: every upcoming game, with the model's probability beside the price of the Kalshi market and the sportsbooks' line.
• Picks: the day's picks most likely to be right.
• Results: final scores, and whether the prediction made before each game picked the winner.

How the predictions are made

Each league has its own statistical model, trained on that league's past games. The models rate every team from its results, weigh its recent form, and take in what is known before the game where it matters — the announced starting pitcher in baseball, for example.

A model is judged only on games it had never seen when it was trained, and its probabilities are calibrated: when it says 60%, the pick should win about 60 times in 100. What a league's model does not take into account is listed on the Predictions page, league by league.

What a probability is, and is not

A probability is an estimate, not a promise: a pick given 65% loses about one time in three. The Results page shows, every day, how the predictions did — the misses with the hits.

The sportsbooks' and the markets' prices are a strong benchmark. Where the model and the market disagree, the market has more often been the one that was right, and the pages say so where it matters.

Where the data comes from

Schedules and scores come from public sources, among them the leagues' own feeds and ESPN. Market prices come from Kalshi and Polymarket, and the sportsbooks' lines from public listings of them. The names and logos of leagues and teams, and the photos of players, belong to their owners and are shown to say who is who.

Who runs iPick

iPick is run by Adrian Guzman, in the United States. It is independent: it is not affiliated with, or endorsed by, any league, team, sportsbook or market named on it.

Questions and corrections are welcome: see Contact.