How it works.
What it costs
to be wrong.
Every edge we surface has a model behind it. Here's exactly what each sport model considers, how confidence tiers are assigned, and what we deliberately don't do.
Basketball model
The NBA model pulls real-time team stats from the NBA Stats API and builds a possession-adjusted win probability for each game. Key inputs:
Baseball model
The MLB model blends pitcher quality, team offense, and park factors into a runs-per-game projection using the MLB Stats API.
Golf model
PGA Tour predictions via the BallDontLie API, factoring course history and recent form to generate winner odds and value picks.
How we size picks.
The edge gap between model win probability and market implied probability determines the tier.
No noise policy.
We never parlay picks. Each game is evaluated independently. Parlays compound variance, not edge.
Teaser adjustments obscure the actual edge. We quote every pick at the straight line.
If a team is "due" but the model disagrees, no pick is issued. Recency narratives are already in the market price.
The record page shows every pick — wins, losses, pushes. The full ledger is public.