Odds come from simulating every remaining game 20,000 times from the real standings and the real schedule as they stood on each date. Team strength is a blend of record and run differential, regressed toward .500. Ties are broken at random rather than by the real tiebreaker rules, so treat a number near a coin flip as a coin flip.

The Top Twenty-Five, Game by Game.

SP+ difference through a normal CDF, sd 16.5, home field 2.5; 20,000 season simulations per club. SP+ is a points-above-average rating, so the difference between two clubs is an expected margin and converts to a win probability at any gap — unlike a rank-difference model, which breaks when the gap is large. Strength of schedule is the mean SP+ of the opponents faced; Notre Dame’s +3.0 against Texas’s +13.7 is most of why their unbeaten chance is so much higher despite a lower rating.

Thirty-Two Clubs, Priced by the Market.

Ratings solved by least squares from 112 published point spreads (residual 0.61 points); implied home advantage 1.53. 20,000 simulated seasons, margins normal with a 13.2-point spread. A published spread is a statement about a difference — the home club minus the away club, plus home advantage — so a hundred-odd of them form a linear system in thirty-two unknowns that solves for a rating per club. We did not impose the home advantage; it fell out of the fit at about a point and a half, which is close to what the modern game is measured at, and is the main reason to trust the rest. Ties are broken at random rather than by the real tiebreaker rules, so treat anything near a coin flip as one.

The Bracket, With Every Season Inside It.

Series odds combine club strength by log5 and then add home field in log-odds, which is the only space where the two adjustments can simply be added: a home edge worth a few points to a .500 club is worth less to a .700 one, and adding it in probability space can push a number past 1. The home edge is a single constant, +0.14 in log-odds, from the .535 home win rate between evenly matched clubs. Strength blends record with the Pythagorean estimate from runs scored and allowed, regressed toward .500 by a prior worth 69 games. Best-of-three is played entirely at the higher seed; best-of-five splits 2–2–1.