College Football 2026
LIVE20,000 simulations of the real 887-game FBS schedule, all ten conference title games and the twelve-team playoff - a ratings model built from three seasons of opponent-adjusted margins, anchored to the AP poll and the title market, predicting every Top 25 game as the season plays out.
points-v3 · AP Week 3 2026-09-13 · updated 2026-09-16 · after 180 games
The race for the national title
Share of simulated seasons each program wins the CFP National Championship.
The twelve-team field12 teams
Five conference champions and seven at-large, seeded straight by committee rank, byes to the top four seeds.
How the field fills
The eleven contested spots - Power 4 + Notre Dame
Show all 14 teams
Green rows mark the eleven most likely qualifiers from this pool; the line under row eleven is the cut.
The Group of 5 bid
The guaranteed spot goes to the highest-ranked of these leagues' champions - the green row is the current favourite to claim it.
The conference title games10 pairings
The likeliest title-game pairing per league, each team's odds of reaching it, and the favourite to win outright.
The week's Top 25 games, called23 games
Every game involving an AP Top 25 team (Week 3 poll), win probability for the home side.
How this is measured
Show all 23 games
Games that matter this week23 games
Points of playoff probability each team stands to swing by winning versus losing, most consequential first.
Show all 23 games
Bubble watch8 teams
Teams most often on the last at-large line: the twelfth team in, or the first team out.
How the model is doing10 games
Against the closing line: 3 of 22. Each pick is frozen the day it is made, days before kickoff. The market keeps moving after that, and the price at kickoff is the best-informed number anyone has. This counts the games where the call we froze early still scored better than that final price (closing Brier 0.1422). Games picked and played between two runs never got a later read and are left out, which is why the count is its own.
Every team, every outcome11 conferences
Expected regular-season wins and the odds of each landing spot, conference by conference.
Show all 16 teams
Show all 18 teams
Show all 16 teams
Show all 17 teams
Show all 13 teams
Show all 14 teams
Show all 16 teams
Show all 18 teams
Show all 16 teams
Show all 17 teams
Show all 13 teams
Show all 14 teams
Show all 16 teams
Show all 18 teams
Show all 16 teams
Show all 17 teams
Show all 13 teams
Show all 14 teams
Get the data: season table as CSV · raw JSON on GitHub
Citizen of Nowhere Picks
Call every game of the week blind, rank your confidence, and take a side on the Upset Radar - the games where this model and the betting market disagree most. The model plays its own card, graded by the same rules.
Play Citizen of Nowhere PicksWhere these numbers come from
Ratings start from three seasons of opponent-adjusted FBS scoring margin (2025, 2024, 2023), anchored to the national-championship futures market and the AP poll, then replayed forward against the real remaining 2026 schedule. Last generated 2026-09-16, after 180 games played. Weekly picks cover AP Top 25 games, blend the model 50/50 with the posted line, and are graded against final scores above; the Brier scores on this page use that same graded record.
How the model works
Each team's rating starts from its opponent-adjusted scoring margin over the last three seasons (2025, 2024, 2023), margins capped, home advantage removed and FCS opponents pooled, then regressed toward the mean (×0.88). On top of the stats sit two anchors, each mapped onto the points scale through the model's own rating-to-title-odds curve: the national-championship futures market (weight 0.35) and the AP poll's vote shares (weight 0.25) - college rosters churn through the portal, so the poll and the market carry roster news that three seasons of margins cannot see. Both anchors fade as real 2026 results fold in. Every remaining game on the actual schedule is simulated as a point-spread probability (home advantage 2.6 points, margin sd 16), each simulated season drawing every team's rating from a distribution (σ 3.5points). Conference standings, all ten title games and the twelve-team playoff play out inside every simulation under the 2026 format: five highest-ranked champions in, seven at-large, straight seeding, byes to the top four. Two stated approximations: conference tie-breaks are record then head-to-head, not each league's full ladder; and the selection committee is modeled as rating plus record - a proxy, because the committee is not a formula. Weekly picks cover AP Top 25 games only, blend the model with the posted line 50/50, and are graded above.
Three tiers of this model run side by side: a stats-only "lite" rating, a "classic" build that adds the title market, and the full "deluxe" build that also folds in the AP poll - deluxe is the production tier shown by default above. Uncertainty shrinks as the season does - the spread of outcomes each simulated season draws from narrows week by week as fewer games remain, and widens back out for a team the stats and the market disagree about most. Each simulated season also draws one correlated error for the whole sport, one for each conference, and one for each team, rather than treating every game as its own coin flip - real seasons run hot or cold together, not independently. Every tier reuses the same random draws for the same simulated season, so a change from one build to the next reflects an actual change in the inputs rather than noise in which season got drawn.