Sports Analytics Demo
NFL Spread and Over/Under Predictor
A chronological model that trains through historical NFL seasons, then backtests against closing spreads and totals without looking ahead.
Current Model
The public demo defaults to Market Blend, which combines the closing line with the chronological team model for lower prediction error while keeping only conservative spread picks. RSM — Experimental shows frozen-model winner, ATS, and separately versioned total projections, with no established betting advantage.
Checking data freshness...
Checking live scoreboard adapter...
NFL Prediction Dashboard
Current model rating state across all teams. Early-season ranks can still reflect prior-season baseline and model strength until more current-season games are played.
| Rank | Team | Strength | Expected Pts | Allowed Pts | Neutral-Field Win Proxy | Rating-Based Playoff Proxy | Injury Adj | Recent Margin Avg (4) |
|---|---|---|---|---|---|---|---|---|
| Loading... | ||||||||
Upcoming Week: All Algorithms
Open the dedicated forecast board for every model and the current computation status.
Matchup Predictor
Select any two teams, enter a market spread and total, and compare the model projection.
Projected total: --
Historical Casino Lines
Select a matchup to view prior regular-season meetings.
| Season | Week | Matchup | Home Spread | Selected Home Line | Published O/U | O/U Result | Favorite Result | Algo Spread | Algo O/U | Final |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading... | ||||||||||
Backtest Results
Results are graded only when the model has a clear enough edge over the published line.
| Season | Games | Spread Picks | Spread Win | O/U Picks | O/U Results | O/U Win Rate |
|---|---|---|---|---|---|---|
| Loading... | ||||||
Chronological Backtest
Games update the model in order, so future results never influence earlier predictions.
Market Comparison
Spread and total picks are made only when the model differs from the betting line by a set threshold.
Research Modes
Market Blend shrinks projections toward the closing line. Mean Reversion uses prior two-year PF/PA anchors plus current-season-to-date form. CS Matrix iterates opponent-adjusted current-season strength with market-line context. Rothstein+ adds a midseason and QB-continuity filter as an active model.
RSM — Experimental
Uses frozen RSM margin and separately versioned Stage 7B total artifacts to show experimental projections. Neither has an established betting advantage.