Indiana enters 2026 as the model's national title favorite.
Action Network built a season-long simulation engine that rates all 138 Division I FBS teams using 12 years of results, play-by-play efficiency data and preseason power ratings, then simulates the rest of the season 20,000 times. Here's what it projects heading into the 2026 season, and how it works under the hood.
2026 National Championship Projections
Run the 2026 season forward 20,000 times and Indiana comes out on top more often than anyone else.
The Hoosiers sit at 16.4% to win the national title, the highest number in the sport, with Ohio State (13.0%) and Texas Tech (11.3%) rounding out a top three that combines for a 40.7% share of the projected championship field.
Top 10 national championship odds, 2026 (pre-Week 1)
| Team | Champion | Runner-Up | Lost Semifinal | Lost Quarterfinal | Lost First Round | Missed Playoff |
| Indiana | 16.40% | 11.20% | 17.80% | 24.50% | 10.20% | 19.90% |
| Ohio State | 13.00% | 9.10% | 14.50% | 21.10% | 10.10% | 32.20% |
| Texas Tech | 11.30% | 11.90% | 20.90% | 33.50% | 10.80% | 11.60% |
| Oregon | 7.20% | 6.60% | 12.10% | 20.10% | 12.80% | 41.20% |
| Notre Dame | 6.90% | 7.80% | 16.20% | 28.20% | 15.80% | 25.10% |
| Utah | 5.20% | 5.60% | 11.80% | 21.20% | 13.50% | 42.70% |
| Georgia | 4.00% | 4.00% | 7.90% | 13.80% | 10.80% | 59.50% |
| Ole Miss | 4.00% | 4.00% | 7.20% | 13.30% | 10.30% | 61.20% |
| Texas A&M | 3.90% | 3.80% | 7.30% | 13.80% | 10.90% | 60.40% |
| Miami | 3.10% | 4.10% | 9.30% | 19.30% | 19.10% | 45.20% |
| Vanderbilt | 2.30% | 2.40% | 5.00% | 9.90% | 9.90% | 70.50% |
| Washington | 2.10% | 2.20% | 4.10% | 8.70% | 8.10% | 74.80% |
| Iowa | 1.90% | 1.90% | 4.40% | 8.90% | 9.20% | 73.60% |
| USC | 1.70% | 1.70% | 3.20% | 7.10% | 6.80% | 79.50% |
| Penn State | 1.50% | 2.00% | 4.50% | 10.20% | 10.80% | 71.00% |
| Oklahoma | 1.30% | 1.40% | 2.90% | 6.10% | 6.90% | 81.30% |
| SMU | 1.10% | 1.60% | 3.70% | 10.50% | 14.60% | 68.40% |
| James Madison | 1.10% | 1.80% | 5.10% | 13.90% | 17.00% | 61.00% |
| BYU | 1.00% | 1.30% | 3.00% | 7.00% | 8.30% | 79.40% |
| North Texas | 1.00% | 1.40% | 3.40% | 9.20% | 12.40% | 72.60% |
Use the tool below to look up your favorite D1 team! Please scroll below:
Texas Tech is the sharpest number on the board. The Red Raiders have the lowest miss-the-playoff rate of anyone outside the top two (11.6%) and the single highest runner-up rate in the sport (11.9%), meaning the model has more confidence in Texas Tech reaching the playoff than it does in Indiana or Ohio State. It just isn't as sold on the Red Raiders winning it all once they get there.
The gap after the top three is real. Oregon and Notre Dame both clear 6% to win it, but their playoff-miss numbers (41.2% and 25.1%) show how much more swings on their remaining schedule than on Indiana's or Ohio State's. Georgia, Ole Miss and Texas A&M are bunched at 3.9-4.0% to win the title, but each still misses the playoff more than 59% of the time in the model's simulation.
Below the top 10, the picture flattens out fast. Only 21 of the 138 teams in the model project at 1% or better to win the national championship, and 106 teams carry a 90% or higher probability of missing the playoff altogether before a single snap of the season. That's the honest range of outcomes a preseason simulation produces: a handful of teams with a real path, a long tail with almost none.
Anyone comparing these numbers against the sportsbooks can check the latest college football betting promos, and traders working outside a traditional sportsbook can look at the best college football prediction market promos.
How the Supercomputer Works
The model runs on five years' worth of layered inputs, all pulled from CollegeFootballData: game schedules and results back to 2014, play-by-play efficiency data (Predicted Points Added, or PPA, which weighs a play by down, distance and field position rather than just raw yards), preseason SP+ ratings, aggregated sportsbook lines and the selection committee's own historical rankings. That's 12 full seasons and roughly 8,600 games used to build and check the engine before it ever touches the current year.
Every team's offensive and defensive rating is refit weekly using only games already played, adjusted for opponent strength. Early in the season, before there's enough current-year data to trust, the model leans heavily on that team's SP+ rating from the year before; that weight starts near total in Week 1 and gets cut in half every six weeks as real results accumulate. A team playing its first-ever FBS season, with no prior rating to lean on, is assigned a below-average starting point rather than a league-average one, calibrated off how every past FBS debut team has actually performed.
Ratings translate to game outcomes through a scoring-margin formula fit once against the full 12-season history, plus a home-field edge worth a little over three points. Every remaining game, and the full 12-team playoff bracket that follows, gets simulated 20,000 times, with randomized variance drawn from the model's own history of real-world upsets rather than an assumed bell curve. That's what produces a range of outcomes for each team instead of a single static number.
The model has been checked for accuracy, not just built and trusted. Games were grouped by the model's predicted win probability into 10-point bands and compared against how often the favorite actually won; across roughly 8,600 historical games, the two track closely across every band. That calibration is what the projections above are built on: informed season-long and week-to-week context, not a standalone wagering signal. There are two other limits worth stating plainly.
The model's stand-in for the selection committee agrees with the real committee's historical rankings 82% of the time, up sharply from 73% using win-loss record alone, but disagreement is most likely for teams sitting right at a cutoff, the last playoff spot or the fourth seed. And every parameter here was tuned and validated against the same 12 seasons it's meant to explain; a stricter out-of-sample test, checking the model against a season it never saw during training, hasn't been run yet.
You can take a look at the dataset behind the model here.
