SEARCH A PICK
Each sport runs through a tuned feature set — what predicts an MLB game has nothing to do with what predicts an NHL game. We don't pretend one model fits all.
MLB uses a Poisson run-scoring engine built on FanGraphs wOBA/xFIP with park factors, early-season prior-year blending, and starter-vs-bullpen ERA splits. NHL rates teams on xGF%/HDCF% differential, goalie GSAx, PDO regression, and rest advantage. NBA ranks on NetRtg/Off+Def Rating gap, pace differential, and back-to-back penalties.
The model is the floor — the market signal is the ceiling. We track line movement from the moment a game opens to the moment it closes: reverse line movement (line moves AGAINST the public side) is the clearest fingerprint of sharp money. We pull live odds across the major US books via Kalstrop, snapshot the open and close, and flag every closing line value (CLV) beat — the only verifiable proof that a model is beating the market, not just winning bets.
This is where it gets interesting. We pull real-time Polymarket sentiment and compare it against sharp sportsbook lines to find sentiment divergence — when the prediction market crowd disagrees with the sharps, someone is wrong, and we want to know who. On top of that, we track 100+ handicappers on the CAPTRACKER leaderboard. When 6+ profitable cappers pile on the same side, that's a consensus signal. When the crowd and Polymarket both fade a team the books are pushing? That's a convergence play — the highest-conviction bets in our system.
support@captracker.markets · DISCLAIMER
