Sport coverage. MLB, NFL, NHL.
NCAA, NBA, EPL, and Golf on the roadmap.
Honest scope. Honest accuracy. Per sport.
We project what we can project well. Each sport runs the same Bayesian-posterior, correlated-MC, ILP pipeline: tuned per sport, with the data sources and training cadence we actually use.
Currently covering
Three sports, fully shipped. Models trained, tracked, audited.
MLB
- Scope: 9 batter components (singles, doubles, triples, HR, BB, HBP, SB, runs, RBI) and 8 pitcher components (IP, K, BB, H, ER, HBP, W, QS).
- Training: batter model Monday night, pitcher model Tuesday night.
- Data: the official MLB stats feed for confirmed lineups and player history; consensus Vegas market for implied team totals; weather and park priors per game.
- Pillar: cross-platform batting-order consensus. #3ESA-verified when the lineup is officially posted, #3platform-projected when only a sister platform has confirmed.
NFL
- Scope: 30 components (26 player components plus 4 DST components).
- Training: Thursday night, ahead of the Sunday slate.
- Data: weekly snap counts and target shares, opponent priors (rush vs pass defense, pace-of-play), weather priors, and consensus Vegas market totals.
- Pillar: full posterior distribution per component. The spread you see is what the model actually believes, not a thousand sims wrapped around a point estimate.
NHL
- Scope: 5 skater components (goals, assists, shots on goal, blocked shots, power-play / shorthanded points) plus 5 goalie components. Forwards and defense fit separately.
- Training: skater model Wednesday night, goalie model Friday night.
- Data: the NHL public api-web feed for rosters, scratches, and matchup state; Vegas market for implied team totals; a rotation-aware heuristic for projected starting goalies.
- Pillar: goalie starter consensus. pESA-verified when DK or FD has flagged the starter, pplatform-projected when only the rotation heuristic has fired.
On the roadmap
Four sports we plan to ship when they're at parity.
NCAA Football
College football is breadth, not depth. 130+ FBS teams, less stable per-game data, and rotations that change week-to-week. We're modeling it, not yet shipping.
NBA
Player-rotation variance is the dominant signal. We will not ship until our minute-projection accuracy is at parity with the other three sports.
English Premier League
International soccer is a different model surface: fewer events per game, different priors, different public data feeds. Worth doing right, not first.
Golf
Tournament golf is a different decision problem. Cut risk dominates portfolio sizing in ways MLB and NHL don't. We're modeling it, not yet shipping it.
When any of these lands, you'll know. We'll publish the same MAE, rank correlation, and top-K accuracy numbers we publish for everything else, and you'll see them inside the product before we put them on this page.