Two Top-12 Teams Lost to Unranked Opponents on Saturday. Both Lines Were Built From the Same Mistake.

Oklahoma scored 51 points against UTEP and did not allow any. Michigan needed a Hail Mary to beat Western Michigan 13-12 and got voted out of the AP Top 25 for winning. One week later the market made Oklahoma a 5.5-point favourite in Ann Arbor, and Michigan won 17-10 without throwing a touchdown pass.

Four hundred miles west, Oklahoma State — 1-11 last season, beaten 24-10 at Tulsa seven days earlier, riding a 21-game losing streak against FBS opposition — beat No. 6 Oregon 39-31 in Stillwater. Oregon closed as a 23.5-point favourite. DraftKings had it at 24.5.

These look like opposite events. A modest home dog covering and winning outright is not the same species as a three-touchdown favourite losing by eight. But both numbers were assembled the same way, out of the same material, and the material was bad.

The direction of the move is the evidence, not the result

The easy version of this argument is that the market was wrong because the favourites lost. That version is worthless. Favourites lose; a 23.5-point favourite is supposed to lose sometimes, and if you only notice the pricing when the underdog wins you have learned nothing except how to read a scoreboard backwards.

The falsifiable claim is about movement, and it does not depend on who won.

Earlier in the week, Oklahoma was a 4.5-point favourite at Michigan with the total at 43.5. By Saturday morning the number was 5.5. Between those two quotes, nothing happened except that people finished digesting a 51-0 win over UTEP and a 13-12 survival against Western Michigan. No injury, no suspension, no weather. The line moved a full point toward the team with the prettier box score, in a game where the two box scores described entirely non-comparable opponents.

That is money reading margin as quality. And margin against UTEP is a measurement of UTEP.

What the Michigan game was actually decided by

None of the inputs that mattered were visible in either Week 1 result.

Michigan won time of possession 33:17 to 26:43. Bryce Underwood — who threw for 111 yards on nine completions, a line that would ordinarily lose a football game — carried the ball 18 times for 87 yards and a touchdown. Michigan committed zero turnovers. Oklahoma committed two. Total yardage went Oklahoma's way, 289 to 263, which is the sort of detail that tells you the Sooners were not outplayed so much as out-possessed and out-protected.

John Mateer, the quarterback whose 225-yard, three-touchdown afternoon against UTEP did the work of moving that line, went 17-of-33 for 189 yards with a touchdown and an interception. Both teams converted almost nothing on third down: Oklahoma 3-of-13, Michigan 4-of-13. The most productive receiver on the field was Oklahoma tight end Rocky Beers, with six catches for 94 yards, in a losing effort.

Michigan scored seven in the first quarter, three in the second, nothing in the third, and seven in the fourth. It is not a performance anyone will put in a highlight reel. It is a performance that was entirely unpredictable from a Hail Mary.

Stillwater is the harder case, and the more interesting one

Oregon's number was not built from one week of data. It was built from a brand, a returning quarterback in Dante Moore, and an opponent whose most recent full season was 1-11 and whose most recent game was a loss to Tulsa. Every one of those inputs was true. The number was still wrong, and the reason is that one of them referred to a team that no longer exists.

Eric Morris is in his first year at Oklahoma State. The 1-11 season and the 21-game FBS skid belong to a previous staff. Markets are reliably slow to reprice a coaching change, because the evidence for a new staff arrives late — you cannot see it until the games start, and by then the number has already been set by the old record.

What Morris's team did on Saturday: 554 total yards to Oregon's 281. Drew Mestemaker threw for 317 yards and two touchdowns and ran for 109 and another. Caleb Hawkins carried 23 times for 110 yards and a score. Wyatt Young caught five balls for 121 and a touchdown. After Moore's 38-yard touchdown to Evan Stewart put Oregon ahead 31-30 early in the fourth, Mestemaker drove Oklahoma State 75 yards and dived over the pile from ten yards out for the lead.

Dan Lanning did not dress it up: "We got our (butts) kicked in a lot of areas. Tons of credit to coach Morris." Morris, for his part, described a roster that "had to listen to how bad they were after hearing how good they were for such a long time."

Kickoff was also pushed back 75 minutes by lightning inside an eight-mile radius of Boone Pickens Stadium, with both teams chased off the field mid-warm-up around 9:40 a.m. local time. A genuine disruption, and one that landed on both sides — we are not going to pretend it explains a 273-yard gap in total offence.

The strongest objection

Two upsets on one Saturday is variance, and a thesis assembled from two data points is not a thesis. This is correct and it is the reason most of what gets written on a Saturday night is worthless by Tuesday. Roughly one Week 2 in every few produces a top-ten team losing to somebody it should beat by four scores; the base rate is not zero, and a market that never let that happen would be badly calibrated in the other direction.

We would also concede that Oklahoma State becoming, per the NCAA, the first FBS team to snap a 10-plus-game losing streak against an AP top-ten opponent is by definition a description of something rare. Rare things are not evidence of mispricing. They are evidence of rarity.

Two things survive the objection. The first is the Michigan line movement, which happened before kickoff and is therefore untouched by the result — money moved a point toward the 51-0 team, and that is a fact about how the market weighted opponent-blind margin, whatever Michigan then did. The second is that the two errors point the same way. If this were variance you would expect the Saturday's surprises to be randomly distributed. Instead both involved a favourite whose price leaned on recent scoreboard margin against weak or non-comparable opposition, and both underweighted things that do not appear in a box score at all: a first-year staff, a quarterback who runs, a home crowd, an offensive line that can hold the ball for 33 minutes.

What would prove us wrong

If Oregon wins out and Oklahoma State loses its next six, the 23.5 was a fair number that met a bad night, and we will have built an argument on a coin flip. If Oklahoma beats a real SEC opponent comfortably in three weeks, the 5.5 was defensible and Michigan simply played its best game.

The version we would actually bet on is narrower and it is available every September: through the first three weeks, a line assembled largely from Week 1 margin is carrying information about schedule strength that it is presenting as information about team strength. Oklahoma State played Tulsa. Oklahoma played UTEP. Michigan played Western Michigan. On Saturday, all three of those facts turned out to be worth less than the market charged for them.