College Football Week 2 Predictions: The Machines Won't Take Texas

Kalshi has winner contracts listed on 120 college football games for the weekend of September 12, and one of them has pulled in more money than any other game with a ranked team in it. Ohio State at Texas, in Austin, at 7:30 on the night of Saturday, September 12. The Longhorns trade around 52 cents, the Buckeyes around 48, and the distance between what buyers are bidding and what sellers are asking is a single cent. Tens of thousands of contracts have changed hands, and the price sits a hair off a coin flip.

Almost every team on this board has yet to play a snap that counts in 2026. USC, which beat San Jose State 42-26 on August 29, is the exception, and Week 1 does not finish until September 7. So the market is doing something genuinely hard right now: pricing 23 games involving ranked teams on last season's evidence, a preseason poll, and an offseason in which 34 FBS programs changed head coaches.

We handed nine AI models the same problem and took the prices away. Each one got an identical fetched fact card and returned a win probability for the home team in all 23 of those games, with no market price, no point spread and no line movement anywhere in front of it. Then each model read the other eight boards, anonymized, and revised once. Laid next to Kalshi, the panel agrees nearly everywhere. Twelve of the 23 land within a point and a half. The disagreements are not spread evenly across the slate, and the loudest one of the week is sitting on the game carrying the most money.

New to event contracts? Start here → how to bet college football on Kalshi.

How This Board Was Built

Every model received the same card, built today from live sources and nothing else. It carried the AP Top 25 currently in force, each team's final 2025 record with points scored and allowed, each team's end-of-2025 Elo rating (a single number for team strength that rises and falls with results, so a 300-point gap means one team was meaningfully better than the other last year), the head-to-head series back to 2019, the venue, kickoff and broadcast, and the Week 1 opponent each team still has to get past before any of this settles. On top of that sat a block of dated reporting the schedule feeds cannot see: the coaching carousel, and who is playing quarterback.

That last block mattered more this season than it usually does. Michigan fired Sherrone Moore in December 2025 and hired Kyle Whittingham on a five-year deal worth an average of $8.2 million a year, ending a 21-year run at Utah in which he went 177-88. Utah promoted his defensive coordinator, Morgan Scalley. Penn State hired Matt Campbell away from Iowa State on an eight-year deal; Iowa State replaced him with Washington State's Jimmy Rogers. Lane Kiffin left Ole Miss for LSU, and Ole Miss elevated Pete Golding to replace him. Those six programs, Michigan, Utah, Penn State, Iowa State, LSU and Ole Miss, appear in six different games on the board below, each of them opening a first season under a new head coach.

Two things about the calendar are unusual enough that the models were told to reason about them out loud. The AP Top 25 on the card is the preseason poll, released August 17, because it is the only poll in force; the first in-season ballot does not land until Week 1 finishes. And Week 1 itself is still being played, so a team can look completely different by the time its Week 2 contract settles. Rankings on this page are context, not evidence, and the models were told not to defer to them where the fetched records said something else.

Each of the 207 seat calls below, nine seats across 23 games, went into our public grading ledger the moment it was made, before any settlement. When the games finish on September 12 and 13, each seat's Week 2 record posts in public, win or lose. That is the entire point of running it this way: the page rebuilds every week of the season on this same URL, and the panel's record follows it around.

The Board: Every Ranked Game On Kalshi

Prices below are Kalshi's midpoint quotes as of September 2, 2026, taken on Wednesday afternoon ET, in cents for the home team's contract, where a price of 74 cents means the market is treating that team as roughly a 74 percent chance to win. "Panel" is the median of the nine seats' home-win probabilities after the revision round. All nine seats on the panel returned a full board this cycle.

A word on liquidity before you read down the column. Ten days out, some of these books are thin: Penn State at Temple has traded fewer than five contracts, Texas Tech at Oregon State fewer than 20, and Arizona at BYU carries an eleven-cent gap between bid and ask. On books like those a quoted price is one market maker's opinion rather than a crowd's, and single-cent moves are noise. The heavily traded games are a different animal.

Game (Away @ Home)	Kickoff (ET)	Kalshi: home	Panel: home	Gap
Florida A&M @ No. 7 Miami	Thu 9/10, 8:00 PM	99.5¢	98%	-1.5
Villanova @ No. 24 Louisville	Fri 9/11, 7:00 PM	92.5¢	93%	+0.5
No. 25 Missouri @ Kansas	Fri 9/11, 8:00 PM	31¢	30%	-1.0
No. 2 Oregon @ Oklahoma State	Sat 9/12, 12:00 PM	13.5¢	3%	-10.5
Howard @ No. 6 Indiana	Sat 9/12, 12:00 PM	99.5¢	99%	-0.5
Arizona State @ No. 8 Texas A&M	Sat 9/12, 12:00 PM	84.5¢	85%	+0.5
No. 10 Oklahoma @ No. 16 Michigan	Sat 9/12, 12:00 PM	49.5¢	52%	+2.5
No. 18 Penn State @ Temple	Sat 9/12, 12:00 PM	10.5¢	5%	-5.5
Western Kentucky @ No. 3 Georgia	Sat 9/12, 12:45 PM	98.5¢	97%	-1.5
Rice @ No. 4 Notre Dame	Sat 9/12, 3:30 PM	97.5¢	98%	+0.5
No. 13 Alabama @ Kentucky	Sat 9/12, 3:30 PM	17.5¢	16%	-1.5
Arizona @ No. 14 BYU	Sat 9/12, 3:30 PM	72.5¢	67%	-5.5
Utah State @ No. 17 Washington	Sat 9/12, 3:30 PM	91¢	92%	+1.0
UC Davis @ No. 19 SMU	Sat 9/12, 4:00 PM	85¢	95%	+10.0
No. 20 Tennessee @ Georgia Tech	Sat 9/12, 7:00 PM	31.5¢	35%	+3.5
Southern @ No. 23 Houston	Sat 9/12, 7:00 PM	98.5¢	98%	-0.5
No. 1 Ohio State @ No. 5 Texas	Sat 9/12, 7:30 PM	52.5¢	35%	-17.5
Louisiana Tech @ No. 11 LSU	Sat 9/12, 7:30 PM	97¢	85%	-12.0
No. 12 Texas Tech @ Oregon State	Sat 9/12, 7:30 PM	9.5¢	4%	-5.5
Iowa State @ No. 22 Iowa	Sat 9/12, 7:30 PM	83.5¢	74%	-9.5
Charlotte @ No. 9 Ole Miss	Sat 9/12, 7:45 PM	98.5¢	98%	-0.5
Arkansas @ No. 21 Utah	Sat 9/12, 10:15 PM	74.5¢	88%	+13.5
Louisiana @ No. 14 USC	Sat 9/12, 11:00 PM	95¢	94%	-1.0

More live boards from the same panel: college football national championship (Ohio State and Notre Dame share the front of the field at 13¢, Texas 12¢, Oregon 11¢) · College Football Playoff odds (Notre Dame 82¢ to qualify, Oregon 77¢) · the Kalshi college football picks hub. Prices fetched September 2, 2026.

Every Seat's Number

Home-team win probability per seat, after the revision round.

Game	Fable	Opus	Sonnet	GPT	Gemini	Grok	GLM	Kimi	DeepSeek	Blend
Florida A&M @ Miami	99	99	97	98	99	97	98	98	99	98
Villanova @ Louisville	94	95	90	93	93	90	93	94	96	93
Missouri @ Kansas	33	32	27	30	28	32	28	30	28	30
Oregon @ Oklahoma St.	3	3	3	4	2	2	3	2	2	3
Howard @ Indiana	99	99	98	98	99	97	98	99	99	99
Arizona St. @ Texas A&M	82	85	87	81	86	85	83	85	85	85
Oklahoma @ Michigan	53	53	52	52	53	52	52	53	52	52
Penn St. @ Temple	5	5	4	6	5	4	6	3	4	5
W. Kentucky @ Georgia	97	97	97	96	97	95	96	97	97	97
Rice @ Notre Dame	98	99	98	98	99	98	98	99	99	98
Alabama @ Kentucky	20	15	15	18	16	16	20	16	15	16
Arizona @ BYU	65	68	68	62	66	67	68	65	67	67
Utah St. @ Washington	90	93	93	90	92	91	90	92	92	92
UC Davis @ SMU	95	96	95	94	96	92	94	95	97	95
Tennessee @ Georgia Tech	40	38	35	39	31	35	35	37	31	35
Southern @ Houston	98	98	96	96	99	96	96	98	99	98
Ohio St. @ Texas	38	35	35	38	30	38	38	33	30	35
La. Tech @ LSU	88	88	87	85	84	81	82	85	85	85
Texas Tech @ Oregon St.	5	5	3	6	4	3	4	2	3	4
Iowa St. @ Iowa	70	74	65	72	85	76	72	78	82	74
Charlotte @ Ole Miss	98	99	97	98	98	97	98	97	99	98
Arkansas @ Utah	87	87	92	88	91	87	88	92	93	88
Louisiana @ USC	94	95	95	93	96	93	92	93	96	94

Every seat on this panel is graded against real market settlements — records to date: Claude Fable 87% on 1,454 graded calls · Claude Opus 87% on 1,526 graded calls · Claude Sonnet 85% on 1,502 graded calls · GPT 85% on 7,696 graded calls · Gemini 86% on 6,384 graded calls · Grok 84% on 5,370 graded calls · GLM 82% on 2,900 graded calls · Kimi 84% on 3,072 graded calls · DeepSeek 81% on 2,944 graded calls. Recomputed daily; the full scoreboard is public.

These numbers are model estimates, not predictions of fact and not financial advice. Kalshi is a CFTC-regulated exchange and its contracts are open to U.S. adults 18+ where it operates; event contracts carry risk and you can lose money trading them.

Every number in this piece gets graded in public once the games settle, and the running tally lives on the model verdict scoreboard. That record also sets the honest frame for everything below: across thousands of graded calls, when this panel and a market disagree by ten points or more, the market has been right roughly two-thirds of the time. A gap is a disagreement worth understanding, not a reason to trade toward the panel.

One Pattern The Board Reveals

Read the gap column top to bottom and the striking thing is how much of it is empty. Twelve of the 23 games land within a point and a half, and those 12 are not the easy ones by ranking. They include Missouri at Kansas, a rivalry game between a No. 25 team and a 5-7 host, and Arizona State at Texas A&M at Kyle Field. What they have in common is money. On the ten most heavily traded contracts on this board the median disagreement is 1.5 points. On the 13 quietest it is 3.5. Where real volume has gone through, the machines and the market have converged on the same answer, which is roughly what you would hope to see from both of them.

The single cleanest example is Oklahoma at Michigan, the noon kickoff on FOX. Kalshi has Michigan at 49.5 cents. The panel's median is 52, and its nine seats landed in a one-point band from 52 to 53, the tightest cluster anywhere on the slate. That is not the panel copying itself; the seats were blind to each other in the first round and arrived there independently. Oklahoma's end-of-2025 Elo was 1810 against Michigan's 1786, a 24-point difference that is inside the noise of a rating like that, and Michigan has the home field. Last season Michigan lost this fixture 24-13 in Norman. Whittingham's first season in Ann Arbor and John Mateer's settled starting job point in opposite directions and stop.

The 50-54 cluster is the correct answer, and the fact that nobody deferred to Michigan's higher preseason ranking is a good sign about the room. — Claude Opus

[Editor's note: Oklahoma is the higher-ranked team here, No. 10 to Michigan's No. 16. The seat's point about the panel pricing the ratings rather than the poll stands; its description of which side the poll favors does not.]

Which brings the pattern to its exception. If liquidity is where the two sides agree, the most liquid contract on this board should be the safest row on the page. It is the opposite, and it is the last section of this article.

Arkansas At Utah: The Second-Widest Argument Of The Week

Kalshi charges 74.5 cents for Utah at home in the 10:15 PM ET window. The panel says 88 percent, the widest gap on the board where the panel is the more confident of the two. The raw numbers behind the models' number are lopsided. Utah finished 2025 at 11-2, outscoring its opponents by 291 points across the season, on an Elo of 1976. Arkansas finished 2-10. The Elo gap is 405 points before home field is counted at all.

The reason the market is not at 88 is almost certainly Morgan Scalley, who is running his first season at Utah after Whittingham left for Michigan, and one detail buried in Arkansas's record. A 2-10 team that was outscored by only 11 points all season lost a lot of close games, which is the profile of a team whose record understates it. Two seats named exactly that tension and landed on opposite sides of it.

Arkansas-Utah should not be a near-lock after Scalley's succession. Utah's 1976 Elo still dominates a 2-10 Arkansas, but first-year staff variance belongs in the mid-80s. — Grok

The counter-argument, from one of the three seats that raised its Utah number in the revision round, is that you can only charge for the new coach once.

I had double-counted, taking a haircut for Scalley's first year AND for Arkansas's -11 differential being better than 2-10 suggests, when the 405-Elo gap plus home field implies mid-90s. — Claude Opus, raising Utah from 84 to 87

Note that the panel's own answer here, 88 percent, is more confident than the market's, which cuts against the lazy version of this exercise. The models are not reflexively fading favorites.

Louisiana Tech At LSU: A Ranked Host The Panel Won't Price Like One

LSU is ranked 11th and costs 97 cents at home against a Conference USA visitor. Eight home-side contracts on this board are priced at 95 cents or higher. Seven of them drew panel numbers within a point and a half of the market. This one drew a 12-point gap, the third widest of the week, and the reason is that the models went and read what LSU actually did last season instead of what its logo implies.

LSU finished 2025 at 7-6, outscoring its opponents by 39 points all season, and its end-of-2025 Elo of 1675 is the second-lowest of any ranked host on this board. Louisiana Tech finished 8-5 with a margin of 85 points, outscoring its own schedule by more than LSU outscored its. The Elo gap between them is 244 points, which is real, and last September in Tiger Stadium LSU won 23-7, which is a grind rather than a rout. Lane Kiffin's arrival is a genuine upgrade the ratings cannot see yet, and every seat priced LSU as a clear favorite. None of them would go anywhere near 97.

LSU at 1675 Elo off a 7-6 season is not a 90-plus host against a Conference USA team that outscored its schedule by more than LSU did. I was guilty of this and corrected it. 94 was me pricing the LSU logo, not the LSU team. — Claude Opus, cutting LSU from 94 to 88

The same seat pointed at the row directly above it as the version nobody is arguing about: Houston is ranked 23rd on an end-of-2025 Elo of 1512, one of the weakest ratings of any ranked team here, and is priced at 98.5 cents purely because it is hosting an FCS opponent that went 2-10. The ranking and the rating disagree about Houston too. Southern is just not the opponent to expose it.

Where The Panel Changed Its Mind

The revision round exists so a model can be argued out of a number in public. The loudest argument inside the panel was the Cy-Hawk game, which produced the widest split between the seats anywhere on the board: 65 at the low end, 85 at the high end, on a game the market prices at 83.5 cents for Iowa.

The card gives both sides real ammunition. Iowa's end-of-2025 Elo was 1937 against Iowa State's 1629, and Iowa is at home. But the six meetings since 2019 were decided by 1, 10, 3, 7, 1 and 3 points, and Iowa State has won three of the last four by 3, 1 and 3.

I had treated the Cy-Hawk game as a rivalry toss-up leaning home, but two peers forced me back to the card: Iowa's end-of-2025 Elo is 1937 against Iowa State's 1629, a 308-point gap before home field, and Iowa State is on its second game under Jimmy Rogers after Campbell left for Penn State. The series history I leaned on reflects Campbell-era Iowa State teams, not this one. — Kimi, moving Iowa from 62 to 78

The seat that moved the other way went to the same list of scores and drew the opposite conclusion.

The card's own head-to-head shows Iowa State has won 3 of the last 4 meetings by 1-3 points each time, directly against a 308-point Elo gap favoring Iowa, a trend I didn't weight in my round-1 reasoning. — Claude Sonnet, cutting Iowa from 70 to 65

And the seat holding the highest number on the board barely gave ground.

Held firm on Iowa State at Iowa at 85 despite the panel averaging around 70. The panel is anchoring to historically tight Cy-Hawk scores but ignoring that a 308-point Elo gap combined with Matt Campbell's departure makes Iowa a much heavier favorite than the historical series implies. — Gemini, which came down from 88 to 85

One seat also surfaced a scheduling detail nobody else had: Louisville plays at Ole Miss on Sunday night, September 6, then hosts Villanova on Friday, September 11. That is a five-day turnaround against the best FCS team on the slate, and it is worth a point or two off Louisville's number.

Ohio State At Texas: The Machines Won't Take The Home Team

Here is the row the whole board has been pointing at. Kalshi prices Texas at 52.5 cents at home, a coin flip with a one-cent spread and more than 56,000 contracts traded on the Texas home-win contract alone, more than a third of the volume across all 23 home-side contracts on this board. The panel puts Texas at 35 percent. That is a 17.5-point gap, the widest of the week by four points, and it is not a shading of the favorite. It flips the game outright: every one of the nine seats has Ohio State winning in Austin, and the most Texas-friendly number on the panel is 38.

The case the models built is short and it is all on the card. Ohio State's end-of-2025 Elo was 2138 against Texas's 1926, a 212-point gap. Ohio State went 12-2 and allowed 130 points across the entire season, outscoring its opponents by 338 points; Texas outscored its own by 132. Julian Sayin, who finished fourth in the 2025 Heisman voting, is back at quarterback. And Ohio State has beaten Texas in both of the last two meetings: 28-14 in the College Football Playoff semifinal at neutral-site AT&T Stadium in January 2025, which closed out the 2024 season, and 14-7 in Columbus in the 2025 opener.

What the models are explicitly not doing is ignoring the venue. Texas hosts this one at night at Darrell K Royal-Texas Memorial Stadium, DKR to anyone who has been there, and the argument between the seats is about how much a night in that building is worth rather than whether it counts. Arch Manning is back as Texas's starting quarterback.

Peers pushing Texas to 0.28-0.33 over-read the 212-Elo gap and ignore DKR at night plus Manning continuity already netting about 0.62 for Ohio State. — Grok, holding Texas at 38, the panel's high

Ohio State's +338 point differential in 2025 and two straight head-to-head wins are stronger signals than home field in a non-conference early-season game. — DeepSeek, holding Texas at 30, the panel's low

The honest way to hold both numbers at once is this. The market has 56,000 contracts of volume behind a price a hair above even money. The panel has read last season's ledger and cannot get Texas above the high 30s. Both cannot be right, and one of the two has a public record: on ten-point-plus disagreements like this one, the market has historically won about two-thirds of the time. September 12 grades it either way, and the result posts to the scoreboard with the panel's name on it.

What Would Change The Panel's Mind

This page re-scores every week of the season, and these are the specific developments between now and kickoff that would move specific numbers, with the direction stated in advance:

• Texas And Ohio State Play Their Week 1 Games On September 5. Texas hosts Texas State at 3:30 PM ET, Ohio State hosts Ball State at 12:30 PM ET. An injury or a struggling offense in either building would move the Austin number more than anything else on this list.
• The First In-Season AP Top 25 Replaces The Preseason Poll Once Week 1 Is Played. Any team that drops out of the top 25 leaves this board, and any team that plays its way in joins it. The composition of the table above is not fixed.
• Jimmy Rogers' First Game As Iowa State's Head Coach, At Home Against Southeast Missouri State On September 5. A convincing performance narrows the panel's 20-point internal split on the Cy-Hawk game toward the low end; a scare widens Iowa's number toward Gemini's 85.
• Alabama Names A Starting Quarterback. The job was an open competition through fall camp between Keelon Russell, the No. 2 recruit in the 2025 class, and Austin Mack. Alabama is a road favorite at Kentucky priced at 83.5 cents, and the panel's five-point internal spread there is almost entirely about this.

When These Games Settle

Date	What settles or what to watch
Thu Sep 3 – Mon Sep 7	The entire Week 1 slate, including every team on this board
After Week 1 Ends, Sep 7-8	First in-season AP Top 25 replaces the preseason poll
Thu Sep 10	Florida A&M at Miami settles, roughly 11 PM ET
Fri Sep 11	Villanova at Louisville and Missouri at Kansas settle
Sat Sep 12 – Sun Sep 13	The other 20 games settle across the noon through 11 PM ET windows
Sun Sep 13	Week 2 grading posts to the model verdict scoreboard

Prices on this page carry their September 2 timestamp, and the board re-scores weekly through the season on this same URL, faster if a major injury lands.

The panel's calls are on the record. Our AI model panel puts a graded, public verdict on the national title race itself, not just this week's games — see the live model verdict on the college football championship for the full field.

Hottest Prediction Markets Right Now

• 2028 Democratic presidential nominee · $210M traded
• 2027 Pro Football Champion · $78M traded
• Pro Baseball Champion · $65M traded
• 2028 U.S. Presidential Election winner? · $64M traded
• 2028 Republican presidential nominee · $62M traded
• Bitcoin price at the end of 2026 · $34M traded

Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.

More on this:

• How To Bet College Football On Kalshi: Markets & Value
• College Football Championship Odds: The Champ Nobody Wants
• College Football Playoff Odds 2026: Notre Dame And The Last Seat
• Kalshi College Football Picks This Week: Every Board, Scored
• AI NFL Predictions Week 1: 4 AIs Vs Kalshi On All 16 Games

The Bottom Line

Twenty-three games, one real argument. On the FCS tune-ups, on the Big Ten and SEC mismatches, and on Oklahoma at Michigan, the one competitive game with heavy volume behind it, the panel and Kalshi are reading the same page: Oklahoma at Michigan is a coin flip, Oregon and Texas Tech are priced as heavy road favorites, and Georgia, Miami, Notre Dame and Ole Miss are priced close to the ceiling. Where the two sides part company, they part over the same question every time. Is a ranking or a rating the better description of a team nobody has watched yet? LSU at 97 cents and Utah at 74.5 are quiet versions of it. Ohio State at Texas is the loud one: the market has 56,000 contracts and a coin flip, the panel has a 212-point rating gap and two straight Buckeye wins over the same opponent, and it will not take the home team. On the night of September 12 one of those two readings gets marked wrong in public, and this page carries the grade.

Hero illustration: OddsShopper, in the house collage style. Darrell K Royal-Texas Memorial Stadium photo by Fredlyfish4, licensed CC BY-SA 4.0; photos cropped, toned, and composited.