Sports covered by Deviate

Follow nine sports. The inputs and available alerts differ by sport.

Market coverage in Live does not always mean a sport has a live win-probability model. The details below distinguish the two.

Baseball

For MLB, the model reads the half-inning, outs, runners on base, and score difference. It compares that state with historical outcomes from decades of play-by-play data. A state needs enough historical examples before it can produce an alert.

Tennis

For ATP and WTA matches, the model reads completed sets and the game score in the current set, with separate historical tables by tour and tournament context. It needs a supported match state and enough comparable history. ITF markets can appear in Live without producing model alerts.

Motorsports

Markets for F1, NASCAR, IndyCar, and MotoGP appear in the Live tab, labelled by series. The current app does not run a live win-probability model for these leaderboard events, so these markets do not produce model alerts.

Basketball

For NBA and WNBA games, the model reads the quarter, time remaining, and score difference. The historical lookup averages over possession, rather than treating possession as a live model input. Other basketball leagues use an estimate based on the lead and time remaining.

Football

For NFL games, the model reads the quarter, clock, score difference, and which team has possession. Possession must be known for an NFL alert. Down, distance, field position, and timeouts are not separate inputs to this historical lookup; other football leagues use an estimate based on lead and time.

Golf

Golf markets appear in the Live tab. The current app does not run a live win-probability model for golf leaderboards, so no golf game-state inputs are used to generate model alerts.

Hockey

The hockey model uses the score lead and time remaining to estimate win probability. It currently uses an estimate formula rather than a measured historical table. Live alerts still depend on a matched game and a gap that passes the alert rules.

Cricket

Cricket markets appear in Live, labelled by format, including T20, Test, ODI, and IPL. Live cricket game state is not currently connected to a win-probability model, so these markets do not generate model alerts.

Soccer

The soccer model reads the match minute, both teams’ goals, and red cards for each side. It compares those inputs with historical results, including the possibility of a draw. A thinly supported state produces no alert.

How the comparison becomes an alert β†’