Why Is LSU Still Rated Highly by Computer Models After Several Bad Games?

BATON ROUGE, La. — LSU’s 2026 season has been a study in contradictions. The Tigers are 5-2, ranked No. 21 in the Coaches Poll, and have looked utterly mortal in consecutive losses to Ole Miss and Kentucky—the latter an overtime defeat to an unranked opponent as a nine-point favorite . By the eye test, this is a flawed team with a shaky offense and a penchant for letting winnable games slip away.

Yet the computers refuse to abandon ship.

After Week 7’s 31-0 shutout of Mississippi State, LSU sits at No. 8 in ESPN’s Football Power Index (FPI) with a 23.0 rating . In ESPN’s SP+ rankings, the Tigers climbed three spots to No. 6 nationally with a 22.6 overall rating . Those are top-ten marks for a team that has beaten exactly one opponent with a pulse—and even that win came against a Mississippi State team missing its star quarterback .

The gap between perception and projection is enormous. Understanding why requires looking at what the models actually measure—and what they deliberately ignore.

The Models Are Not Polls

The first thing to grasp is that FPI and SP+ are not voter rankings. They are predictive systems designed to answer a forward-looking question: how good is this team, really, and how will it perform going forward?

That means they weight underlying efficiency metrics—success rates, explosiveness, opponent adjustments—far more heavily than raw wins and losses. One ugly afternoon against Kentucky does not erase a body of work that, in the models’ eyes, profiles as one of the nation’s most efficient teams.

And LSU’s efficiency profile is genuinely strong. Through five weeks, the Tigers were averaging 43.6 points per game while allowing just 15.2 . Their first-team defense had surrendered one second-half touchdown. Their run game had already scored nearly twice as many rushing touchdowns (19) as it had in all of 2025 (10) . SP+ ranked LSU’s defense No. 3 nationally and its offense No. 16 . Those are not the numbers of a mediocre team.

The Schedule Argument Cuts Both Ways

Here is where LSU’s case gets complicated—and where the models may be getting fooled.

LSU’s dominant early-season statistics were compiled against a schedule that has been, charitably, soft. The Tigers blew out Clemson and Texas A&M by a combined 86-16, but both programs have been below their usual standard in 2026 . Louisiana Tech and McNeese State were guaranteed wins. The only quality opponent LSU faced in the first five weeks—Ole Miss—beat the Tigers and led by 17 at halftime .

That context matters. When SP+ and FPI adjust for opponent strength, they are supposed to account for this. But no model is perfect. The efficiency numbers LSU posted against overmatched competition carry weight in the algorithms, and that weight does not evaporate simply because the competition was weak.

The models are also projecting forward. FPI’s remaining schedule projections gave LSU a 76.8% chance to beat Mississippi State, a 71.3% chance at Auburn, and—notably—a 54.4% chance to win at Tennessee . The algorithm sees a path to nine or ten wins, even with Alabama and Texas looming in November.

The Kentucky Loss: A Feature, Not a Bug

If there is one result that should have cratered LSU’s computer ratings, it is the Kentucky game. The Tigers held multiple leads as a nine-point favorite and could not close the deal in overtime . That is the kind of loss that screams “fundamentally flawed team.”

Yet FPI actually improved LSU’s rating after the Mississippi State win, moving it from where it stood at the end of Week 6 . The Kentucky loss, in the models’ view, was an outlier—a close game decided by a handful of plays rather than a systemic collapse. Predictive systems are designed to smooth out variance. They do not overreact to single-game results the way humans do.

That is both the strength and the weakness of the approach. The models are right that one overtime loss does not define a season. But they are blind to the pattern—the second-half comebacks that fell short, the inability to put away inferior opponents, the sense that this team is less than the sum of its parts.

The Eye Test Problem

College football fans and analysts have noticed the disconnect. After FPI placed LSU at No. 8 earlier this season, Dan Wolken of USA Today posted on X: “If your metric has LSU at No. 8 and Oregon at No. 9, adjust the formula” .

That sentiment captures the frustration. The models can see that LSU’s defense is elite by efficiency metrics. They cannot see the look on the sideline when Kentucky converts a third-and-long in the fourth quarter. They can quantify LSU’s explosive-play rate. They cannot quantify the cumulative psychological weight of consecutive losses in a season that began with College Football Playoff expectations.

The selection committee, which will release its first rankings in November, will watch the games. It will see that LSU’s two losses came against teams the Tigers were favored to beat—or at least expected to compete with. It will see a team that has not beaten a ranked opponent at full strength. And it will care far more about those realities than about an SP+ rating that says LSU is the sixth-best team in America.

What the Models Are Really Saying

The computers are not saying LSU is fine. They are saying LSU is talented—that its efficiency profile, adjusted for tempo and opponent, still projects as one of the nation’s better teams.

But “better” is not the standard in Baton Rouge. The Tigers have a brutal second-half schedule: Auburn on the road, then Alabama and Texas at home, then a trip to Tennessee . FPI gives LSU just a 38.6% chance to beat Alabama and a 42.7% chance to beat Texas . If the Tigers lose either of those games, their playoff hopes are effectively dead.

The models will keep LSU ranked highly as long as the underlying numbers hold up. But the numbers are running out of time to matter. LSU has six games left to prove the algorithms right—and the eye test wrong. The first test comes Saturday at Auburn. The real ones come in November.

The machines are patient. The committee will not be.