Detroit Outscored Its Opponents by 55 Runs. It Is Twelve Games Under.
The Tigers have scored 641 and allowed 586. A club that outscores people like that wins about 79 games. They have 67. The whole gap sits in one column of the schedule.
The question: can a team outscore everybody it plays and still finish twelve games under .500?
Yes, and Detroit is doing it right now. They have outscored their opponents by 55 runs and sit at 67–79.
Why that is surprising. Runs are how you win. A club that scores more of them than it allows, over 146 games, has essentially never finished this far under. Detroit is not slightly unlucky; it is close to the limit of the thing.
What you get: a way to tell which parts of a record will still be there next year.
Where the twelve wins went
Nowhere mysterious, and we did not have to dig. Look at the close games.
Detroit is 14–27 in one-run games. That is a 34% win rate in the games decided by a single run, against a league that by definition wins half of them. They are 5–9 in extra innings on top of it.
So the season has a shape: win the ordinary games comfortably, lose the tight ones almost every time. The runs pile up in the wins and the losses are all narrow. The standings only count how many, never by how much.
Only four clubs in baseball are outscoring their opponents and losing more than they win — Detroit at +55, Pittsburgh at +26, Miami at +10, Washington at +8. Detroit’s gap is double the next one.
So is Detroit bad at close games, or unlucky?
This is the whole question, and it is answerable.
If winning one-run games were a skill — a bullpen, a manager, a bench — then clubs good at it one year would be good at it the next. That is what a skill does. So we took every club from 2021 to now and compared each season’s one-run record with what that same club did the following season. We ran it twice, on two different things, so we would have something to compare against.
On the right is our control: overall winning percentage, one season against the next. It correlates at +0.57. Good clubs stay good; that is a skill, and you can see the line in the cloud.
On the left is the one-run record. It correlates at −0.12 — about 1% of next season explained, and pointing the wrong way. There is no line. There is a blob.
Being good carries over. Winning the close ones does not. They look identical in a standings table and they are not the same kind of fact.
The ProfessorWhat that means for Detroit
Their 14–27 tells you almost nothing about next April. It is not a diagnosis; it is a coin that came up tails.
Be careful with the other half, though. Nothing here says Detroit is secretly a 79-win club now. They have 67 wins and those are real; nobody gives them back. What the runs say is that the process underneath produced better baseball than the record shows, and the process is the part that shows up again next year.
The record is what happened. The run differential is the better guess at what happens next. Both are true and they are answering different questions.
What to take home: ask which part of the record repeats
Any summary of a year mixes two things — what you did, and what happened to you. They arrive fused and they behave completely differently.
The test we used above separates them, and you can run it on almost anything: does this measure predict itself next time? If it does, it is a property of the thing. If it does not, it was weather.
A sales team that hit target three quarters running is different from one that hit it once by a hair. A fund with a decade of returns is different from a fund with a good year. A surgeon with a thousand cases is different from one with forty. In each pair the headline number can be identical, and only persistence tells you which is which.
So when somebody hands you a record, ask what share of it will still be there next year. For Detroit’s close-game collapse, the answer is about one per cent.
Baseball is ranked by championship leverage: we simulate the rest of the season 60,000 times, then split those seasons by who won each of today's games. College football cannot be simulated that way, so it is ranked by how close to a coin flip the ratings make it — a lopsided game teaches you nothing. Tomorrow this block reports what happened to today's pick.
Notes & sources
All figures from the MLB Stats API, pulled 9 September 2026. Detroit: 67–79, 641 runs scored, 586 allowed, a differential of +55. One-run games 14–27; extra-inning games 5–9.
Expected wins use the Pythagorean estimate with an exponent of 1.83, the standard fit for modern run environments: 6411.83 / (6411.83 + 5861.83) = .541, which over 146 games played is 79.0 wins against an actual 67. The shortfall is 12.0.
Clubs outscoring their opponents while under .500, same date: Detroit +55, Pittsburgh +26, Miami +10, Washington +8. Pittsburgh is 24–17 in one-run games, so whatever explains their gap is not this.
Persistence test: every club appearing in consecutive seasons from 2021 through 2026, giving 150 club-season pairs. One-run winning percentage correlates with the next season’s at −0.117 (t = −1.43), an r-squared of 1.4%. Overall winning percentage, identical clubs and identical method, correlates at +0.567, an r-squared of 32.1%. The second figure is the control that makes the first mean something.
Two limits. 2026 is incomplete, so its pairs use a partial season and will shift slightly by October. And a correlation near zero establishes that one-run records do not repeat on average; it cannot rule out a particular club having a genuine and lasting problem in close games. It says only that a single season’s record is not evidence of one, which is what the claim usually rests on.
