Full Data Stack Lab · charts Charts home Rankings Scorecard Matchup Model Season Tournament Dashboard Headline numbers for the latest season, and the way into the six boards that carry the detail. 2026 Season 5,783 Games modelled 72% Best honest forecast 0.178 Brier score 2022202320242025202670%75%AccuracyThe number the lab exists to moveElo's regular-season accuracy, scored only on games it had not seen The top tenBy adjusted efficiency margin, which is the model's own opinion#TeamConferenceRecordMargin1Michigan WolverinesBig Ten36-3+41.92Duke Blue DevilsACC34-3+40.53Arizona WildcatsBig 1235-3+37.94Illinois Fighting IlliniBig Ten26-8+36.65Florida GatorsSEC26-8+35.16Houston CougarsBig 1229-7+34.57Iowa State CyclonesBig 1227-8+33.78Purdue BoilermakersBig Ten28-9+32.99Gonzaga BulldogsWCC29-4+31.610UConn HuskiesBig East33-6+30.3 These boards are YAML. Each one is a file in the same repository as the dbt models it reads, so a chart and the column behind it change in one commit and get reviewed together. The pipeline is the same one behind the Evidence dashboard, which the last button above goes to: public APIs into Parquet, DuckDB and dbt in the middle, Dagster on a nightly schedule. These pages are a second opinion on the presentation layer, not a second pipeline. Built with dbt Charts. The source for every board is in full-data-stack-lab. Data as of 01:14 UTC on 16 Sep 2026 made withdbt Charts