Saturday night, September 12: where the model and the market disagree

Twenty FBS games kick off at 7:00 pm ET or later tonight. For each one the site shows the market's price beside our model's win probability. This note lists the games where those two numbers are furthest apart, and the ones where they agree.

The marquee

Ohio State at Texas, 7:30 pm ET. Our model gives Texas a 47 percent chance to win. Kalshi's last traded price has Texas at 54 percent, on 2.9 million dollars of volume, by far the most traded game on the board. The model's projected margin is Ohio State by 1.2 points with a projected total of 50. Both numbers say the same thing: this is a coin flip, and the market leans slightly more toward the home side than we do.

The widest gaps

Home-side win probability, model versus Kalshi:

• 7:00 pm, Tulsa at Sam Houston: model 47%, Kalshi 17%, gap 30 points
• 7:15 pm, San Diego State at UCLA: model 61%, Kalshi 81%, gap 20 points
• 7:30 pm, Navy at Florida Atlantic: model 17%, Kalshi 35%, gap 18 points
• 7:00 pm, Tennessee at Georgia Tech: model 35%, Kalshi 18%, gap 16 points
• 7:00 pm, Middle Tennessee at Marshall: model 70%, Kalshi 81%, gap 11 points
• 7:00 pm, South Alabama at Tulane: model 86%, Kalshi 76%, gap 10 points

Tulsa at Sam Houston is the largest disagreement on the slate. The model sees a one-point game. The market sees Tulsa as a four-to-one favorite. When we validated this model against the market across sixty live games last week, only two disagreed by more than fourteen points, so a thirty-point gap is rare. The usual reason is information the model does not carry: it is a ratings model built from results, and it knows nothing about injuries, transfers or a quarterback change. Treat the gap as a question about what the market knows, not as a number to act on.

Where they agree

Iowa State at Iowa (model 84 percent for Iowa, Kalshi 85), Louisiana Tech at LSU (93 versus 99), Bowling Green at Nebraska (92 versus 97), Georgia Southern at Clemson (85 versus 91), Southern Miss at Auburn (93 versus 97), Arkansas at Utah (82 versus 81), Louisiana at USC (97 versus 98). Seven of tonight's twenty games sit within seven points of the market.

MLB tonight

Five games start at 7:00 pm ET or later. On every one of them the model's home win probability and the moneyline's implied probability are within one point of each other: Cincinnati at Milwaukee 68 versus 69, Philadelphia at Atlanta 58 versus 59, White Sox at St. Louis 52 versus 52, Texas at Arizona 57 versus 57, Seattle at Athletics 42 versus 42. No disagreement to report.

Method

The CFB number is a moneyline win probability from a ratings model fit on FBS-versus-FBS games only. Calibrated on 4,478 out-of-sample games from 2020 to 2025, its stated probabilities land within about two points of the realized rate on average. We do not publish spread or total probabilities for CFB, because both measured worse than a coin flip in that test. Kalshi prices are the last traded price for the home side's "yes" contract as of 6:41 pm ET and move until kickoff. Games against FCS opponents carry no model number.

Nothing here is a recommendation. Nothing on the site is.

Every game on the slate: troxler.io/games

The public record, losses included: troxler.io/record