The model starts with power ratings, simulates the season and playoff thousands of times to turn those ratings into title win probabilities, and then reconciles the output against de-vigged market prices to produce a single fair value per team. The result is the national-title board: Notre Dame at 11.7%, Ohio State at 11.3%, and Oregon at 10.9% lead the field.
That number, the fair value, is the anchor for everything else on the desk. It is what the model believes a contract is worth before venue and vig enter the picture.
Why start with power ratings?
Power ratings compress each roster, schedule, and returning-production signal into a single strength number. They are the raw material because they answer the only question a title model actually needs to solve: in any given matchup, how often does team A beat team B.
A rating on its own is not a probability. Ohio State sitting above most of the field does not directly say how often the program wins the title; it says how often it should win a neutral game against a given opponent. The simulation is what closes that gap.
How do power ratings become a probability?
The model runs the full season and the twelve-team bracket many times over. Each simulated game is decided by the rating gap plus a margin for variance, so upsets happen at roughly the rate they happen in reality. Count how often a team survives to lift the trophy, divide by the number of runs, and the raw title probability falls out.
That step is why a strong rating does not always translate into a strong title number. Path matters. Indiana's 8.1% and Miami's 7.1% reflect not just team strength but the conference and bracket gauntlet each has to clear. The simulation prices the road, not only the roster.
Why strip the vig from market prices?
Raw contract prices across every title team sum to well above 100% because each venue bakes in a margin. Comparing a single quote to the model without correcting for that margin overstates every probability. The model de-vigs by rescaling the full field so the implied probabilities total 100%, then blends venues into one consensus.
This is the difference between fair value and best price. Fair value is the de-vigged consensus number; best price is the cheapest contract on offer across Kalshi and Polymarket. Notre Dame's 12c best price on Polymarket and its 11.7% fair value describe two different things: the cost of entry and the model's probability.
Venue shopping is where that split pays off. For traders comparing venues, promo access such as Kalshi FADE or Polymarket TGSWC changes net cost, but it does not change the underlying fair value the model computes.
Where do model and market line up now?
Across the top of the board, fair value and best price track closely, which is expected in a liquid market weeks out from kickoff. Notre Dame, Ohio State, Oregon, and Texas all price within roughly a cent of the model's number.
The exception worth flagging is Georgia: a 6.3% fair value against a 5c best price on Polymarket. That is the widest spot where price trails the model on the current board. It is a signal to examine, not a verdict, because both the model and the market can be wrong and prices reprice on news.
How to read the fair value board
The workflow is consistent across every market the desk tracks: power ratings feed the simulation, the simulation produces probabilities, de-vigging turns raw quotes into a comparable consensus, and the gap between fair value and best price highlights where to look. Georgia's 5c print against a 6.3% model number is exactly the kind of gap the process is built to surface.
None of this is a recommendation. The model is a probability estimate, prices reflect real liquidity and information, and the two converge and diverge as the season approaches. The board is a research tool for reading CFB futures as probabilities, not a call on any single contract.
