Method
Subject-resolved tossup-contest engine (Dixon-Coles/KenPom analogue) blended 0.7 with market-implied ratings. Ratings are jointly inverted from BOTH the conference-win and top-4 futures via round-robin simulation. Single-game outcome SD ~56 pts.
Caveats
- Week 1: no NSBA4 game data yet. Win probabilities are intentionally modest (0.20-0.65) because backtest shows roster talent explains only ~6% of single-game margin variance.
- Totals anchored to NSBA3 regular-season scoring (mean 160); 'solvable by majority' packets could push higher. Total is the lowest-confidence number.
- Spreads are large because NSBA games are high-variance blowouts (margin SD ~57).
- Rosters modeled at full strength (no availability adjustment).
- Home advantage = home team removes one toss-up/bonus pair of the away team's best subject (no separate baseline home-points edge).
- Top-4 futures revealed the conferences are far more balanced than the conference-win futures implied (e.g. ChessFun/cryo are competitive, not cellar teams), so spreads are modest and no morale-shading is applied.
- West conference-win and top-4 markets disagreed (esp. cryo, czz); ratings are a joint hedge between them.