How the models work

Per-game logit fit on 72,809 D1-to-D1 games from 2014 through 2024, anchored to current PEAR Ratings. The rating-diff slope is K_DIFF = 0.37 (95% CI [0.364, 0.379]). Home-field advantage is HFA_LOGIT = 0.23, roughly a 6% bump (CI [0.217, 0.249]). Per-game Gaussian rating noise is σ ≈ 2.0, which captures the variance you get from starting-pitcher identity. Fatigue costs 0.40 rating points per game already played in a regional or super series. Bracket sanity check: across 11 historical NCAA tournaments the host-to-super advancement rate was 61.3%, which lines up with these constants better than the 25-year literature anchor of 68.8%.

Dual-Elo system. Each team carries a regular-season Elo and a playoff-only Elo; the two get blended for postseason matchups. In-series state (1-0, 2-1, etc.) modifies win probability on each subsequent game because trailing teams play differently. Rotation ratings shift live for injuries. The bias correction is tuned against 2024 and 2025 playoff outcomes -- 166 backfilled games. The playoff-totals correction is learned live by a nightly self-tuning loop.

12-team CFP simulator. Committee-proxy seeding is learned from 2014–2024 CFP history, so the model picks the kind of teams the committee actually picks. First-round games are on campus, with home-field worth 2.5 points; QFs, SFs, and the Final are at neutral sites (no HFA). Deterministic mode picks the favorite at every node. Sampled mode adds per-game Gaussian noise σ pulled from the ratings JSON, so a single sampled bracket walk has upsets baked in.

The current NFL prediction stack. A long-running team-minus-QB Elo rating updates after each game from margin of victory (FiveThirtyEight-style); a per-player QB Elo travels with the quarterback and gets updated from an independent statistical composite (passing EPA + CPOE + rushing share) opponent- adjusted for defensive quality; a situational layer adds rest, weather, primetime, divisional, stadium-HFA, and injury aggregates. 2025 holdout Brier within ±0.002 of FiveThirtyEight's published 2025 number. Full methodology, calibration, and the six Phase 9 findings (including a face-validity rebuild story) at /methodology/nfl-v6.

AUDIBLE is the NFL stack: six position rankings (QB, RB, WR, TE, OL, DEF) plus a team-strength composite built from eight callsigns (POCKET, ROUTE, HINGE, ENGINE, TRENCH, STORM-PASS, STORM-RUSH, BOOM) with Phase B Ridge-fitted weights. Every published ranking ships with a within-career rolling-8 persistence baseline so you can see exactly where the model adds information. Two of six positions clear the bar: QB at +0.25 R² and TE per-game at +0.29 R². The other four (RB, WR, OL, DEF) are honest descriptive rankings. Read the full breakdown at /methodology/audible and see the live table at /nfl/players.

We anchor every projection to a composite of the major preseason sources (Sleeper, ESPN, FantasyPros, averaged at the component-stat level) and do not claim an accuracy edge on top: tested cleanly across six held-out seasons, the expert consensus ranks players better than our own model (0.70 vs 0.60 within-position rank correlation, every year), and blending our model in does not help. What we add is what the consensus leaves out. Every player ships a floor-to-ceiling range from 200 correlated season simulations, and that range method is calibrated (the 80% range held the real outcome about 80% of the time on held-out seasons). Rankings are value-over-replacement for your exact league, and we price per-game bonuses. An earlier version of this entry cited a backtest beating consensus and ADP; that rested on a contaminated historical feed and was retracted. The dated correction, what survives, and the misses at /methodology/nfl-fantasy and the live board at /nfl/fantasy.

Betting tier: the team-strength rating itself is Tier 1 — descriptive (see the Team Power page); the win-total / over-under comparison below is Tier 2 — market-tracking: a calibration check that the model agrees with the line, not a bettable signal.

Off-season-adjusted team ratings with QB and coach overrides layered on top of a prior-year base rating. Calibrated against DraftKings 2026 win totals: 25 of 32 teams land within 1.5 wins of the market line, with a median miss of about 0.8 wins. The team-by-team breakdown lives in our internal docs/over_under_vs_vegas_predictions.md (labeled Tier 2 — calibration, not tradeable picks); we are cleaning up a public version for the regular-season run-up.

Ensemble forecast pulling NOAA, Open-Meteo, GFS, and ECMWF among others. Each member gets bias-corrected per city, the result is converted to probability bands for daily high and daily low. Two known calibration regimes are in our history: pre-EMOS and EMOS v1. The live system runs EMOS v1 with a confidence cap above 30% predicted probability, applied after a 12K-game backtest showed the model gets overconfident in that range. Calibration tables ship to /calibration as resolved bets accumulate.

The misses. Every published rating clears a calibration check: we bin actual outcomes by predicted probability and see whether the line of best fit is y = x. Where it isn't, that is a model bias we either correct or own. Calibration tables live at /calibration as each sport accumulates enough resolved samples.