College basketball moneyline picks generated from a quantitative model focused on home-underdog edges, tempo-adjusted efficiency, conference matchup dynamics, and market mispricing.
College basketball has more than 350 Division I teams and ~5,000 regular-season games. That volume is a feature, not a bug — but only if you can systematically separate the games where the market is pricing accurately from the games where it is not. The V19a CBB model is built around one specific structural inefficiency: home-underdog moneyline mispricing in matchups where the home team's tempo-adjusted efficiency profile is closer to the visiting favorite than the line suggests.
Each candidate game is scored on a fixed grid of efficiency, pace, conference context, and travel/rest factors. Plays are tier-sized to 1u, 2u, or 3u based on edge strength. The model fires roughly 107 plays per season — selective by design — and only when the home dog's fair win probability beats the moneyline's implied probability by a margin that clears vig and absorbs variance.
College basketball edges are smaller than the public assumes but more durable than the casual model gives credit for. The V19a model focuses on five pillars:
Raw points-per-game lies. The model uses adjusted offensive and defensive efficiency per 100 possessions, opponent-strength weighted, as the core team-quality measure. A team that scores 80 against weak defenses isn't the same as a team that scores 75 against top-50 defenses.
Two efficient teams playing at slow tempo produce a different fair total — and a different variance profile — than two average teams playing fast. The model encodes both teams' true pace and reconciles them against the matchup's expected possession count.
Home-court advantage in CBB is real but wildly non-uniform — a Pac-12 mid-tier on a weeknight is a different home environment from a Big East rivalry at a 12,000-seat arena. The model carries a program-specific home premium rather than a flat home boost.
Conference rematches, late-season fatigue, and tournament-stakes context all shift fair lines in ways the market is slow to adjust to. The model encodes these conference-phase effects directly into the projection.
The model only fires when its fair home-dog win probability beats the moneyline's implied probability by enough margin to clear vig and meet a per-cell ROI-sized tier threshold. The gap — not the upset narrative — is the entire game.
Plays are released as opening lines settle the morning of the game. Conference play is the most plentiful window — January through early March is where most plays fire. Postseason coverage runs through Selection Sunday, the conference tournaments, and into the NCAA Tournament.
The model is calibrated to fire only on real home-dog edges, not to fill a daily quota. Some nights produce three to four CBB plays, some produce zero. Volume averages roughly 107 plays per full season — selective by design.