The model turns power ratings into fair value in three steps: rate every team, simulate the season thousands of times to convert those ratings into a title probability, then line that probability up against de-vigged market prices. The output is a single percentage per team. On the current national title board, that process puts Texas, Oregon, Notre Dame and Miami at 8.5%, Ohio State and Indiana at 6.9%, Georgia at 5.2% and LSU at 4.4%.
How do power ratings become a win probability?
A power rating is only a relative strength number. It becomes useful when two ratings are placed on either side of a game and translated into a margin, then into a win probability for that single matchup. Home field, opponent quality and schedule all enter here, so a strong rating against a brutal slate does not automatically produce a high title number.
The season is then simulated many times. Each simulated run plays out every game, resolves the conference races and the bracket, and records who finishes as champion. Run it enough times and the share of runs a team wins becomes its raw title probability. This is why the model can separate a team that is rated highly but schedule-blocked from one with a cleaner path.
What turns a raw probability into fair value?
Raw simulated probabilities are not directly comparable to screen prices, because listed contract quotes include margin. To make the comparison honest, the market side is de-vigged: the full field of prices is normalized so the implied probabilities sum toward 100% rather than the inflated total a raw board carries. The result is a market-consensus fair value that can be set beside the model's simulated number.
Where the two agree, the board is efficient and there is nothing to do. Where they diverge, the gap is the signal. The model's job is not to be louder than the market but to flag the specific contracts where its simulated probability and the de-vigged consensus disagree enough to matter.
Why do the top four all read 8.5%?
The four-way tie at the top is the model refusing to fake precision. Texas, Oregon, Notre Dame and Miami all resolve to 8.5% because the simulated season places them in one tier with paths of similar quality. Small rating differences between them are inside the noise of a season that has not been played, so the model reports them as equals rather than inventing a ranking.
Below that tier the spacing widens in a readable way. Ohio State and Indiana at 6.9% form the next step, Georgia at 5.2% sits alone, and LSU at 4.4% closes the group shown here. Each drop reflects a thinner slice of simulated championships, not a subjective seeding.
Where does price diverge from fair value?
Best prices sit above fair value across the board, which is the expected footprint of margin in raw quotes. Texas is 8.5% fair value against a best price of 11c on Kalshi; Ohio State is 6.9% against 9c; Georgia is 5.2% against 7c; LSU is 4.4% against 6c. The consistent gap between the percentage and the cent price is the vig the de-vig step strips out.
That is the whole point of putting fair value next to best price: it shows how much of a quote is probability and how much is margin. For readers comparing venues, the FADE code on Kalshi is one entry point, though the model treats the price itself, not the promo, as the variable that decides whether a contract is worth attention.
What the fair value number does not promise
Fair value is a modeled estimate, not a forecast of the season's result. It compresses thousands of simulated outcomes into one number, and both the ratings that feed it and the market it is checked against can be wrong. A team at 8.5% loses the title in the large majority of simulated runs; the figure describes a distribution, not a destiny.
The model earns its keep over many contracts and many seasons, not on any single ticket. Treated that way, the pipeline from power ratings to fair value is a lens for spotting where price and probability part company, and nothing more than that.
