Ninety-One Chances. Six Upsets.

Is this season really stranger than usual, or does it only look that way? The Athletic reported this week, from TruMedia, that unranked teams have beaten ranked ones just six times through Week 5. That is tied with 2023 for the fewest since 2019.

We went and counted. The six is right. We get six too, using the AP Top 25 a team actually carried into the week, and counting only wins by teams that play in the same division.

But the comparison is doing the wrong job. “Fewest since 2019” is a rank, and a rank hides the thing you want to know: fewest out of how many tries? Here is the number nobody printed. There were ninety-one chances at an upset this season, more than in any year since 2019 — and the favorites held anyway. By the end of this you will know why that distinction matters every time somebody shows you a record.

Look at 2022 and 2026. Both offered ninety-one games with one side ranked and one not. 2022 produced fourteen upsets. This year produced six. Identical opportunity, less than half the result.

How Strange Is Six?

Count the seven seasons before this one and the unranked side won 80 of 576 such games — 13.9% of the time. That is the base rate, and it is the only honest yardstick here.

Apply it to this season's ninety-one games and you would expect about 12.6 upsets, give or take 3.3. We got six.

That is two standard deviations below normal, or about a one-in-forty-five season. Not a freak and not a glitch; rare, in the sense of the sort of thing you might see twice in a lifetime of Saturdays. You are watching one now.

Where Our Count Parts Company

Being straight about this, because we checked their number and ours only matches on one of the four.

Use the same rule that reproduces their 2026 figure and their older seasons come out higher for us. We count 17 in 2021, not 14; 7 in 2023, not 6; 9 in 2025, not 8. So on our count this season is not tied with 2023 at all. It stands alone at the bottom.

We tried to find the difference and could not. Swapping the Coaches poll for the AP brings 2021 to fourteen. But then this season becomes ten, and their headline falls apart. Trimming to four weeks fixes 2023 and breaks 2021. No single rule we tried reproduces all four of their numbers at once, so we are not going to invent a reconciliation and pretend we found one.

None of this touches their point. Six is six, and six is the fewest.

The Trap We Nearly Fell Into

One thing worth showing you, because it is the sort of error that produces a confident wrong answer.

To call a game an upset you need to know who was ranked before kickoff. Polls come out mid-week; a database filing one under “week 5” might mean the poll that opened that week, or the one that closed it. Use the wrong one and every upset is graded against a poll that already knows the result. The ranked team has dropped out, so the upset disappears.

We checked rather than assumed. Virginia entered week three ranked 25th, lost that week, and was gone from the following poll. So the poll filed under a week is the one teams carried into iti, which is the one we want.

What to take home

When a number arrives as a ranking — fewest since, most since, worst in a decade — ask what it is a fraction of. The count alone cannot tell you whether the world changed or whether there were simply fewer chances for it to.

Here there were more chances than usual and fewer upsets, which is the harder and more interesting fact. Favorites are not merely winning; they are winning on a card that gave them more room than usual to slip.