Twenty-Six Baseball Franchises. Not One of Them Runs in Cycles.
Every fan believes their club rises and falls on some hidden clock — a good decade, a lean decade, a rebuild, a window. It is the most natural story in sports and it is not in the data. I tested all twenty-six long-lived franchises against a proper null and found zero real cycles. What is there instead is simpler, and you can use it. Every franchise has a level it returns to, and a speed at which it returns.
The question: does your team really move in cycles — a good decade, a lean decade, a window that opens and shuts?
The answer: no. I tested all twenty-six long-lived franchises against a fair comparison, and the clock is not in the data.
Why that is surprising. Everyone can name their team’s cycles. The trap is that a casual look will always find one; any series with a little year-to-year momentum looks wavy to the eye.
Here is the belief, stated plainly enough to test. A franchise is supposed to move in waves: it climbs for some years, peaks, falls, and comes back.
How you actually test for a cycle
The trap is that we will always find one if we look casually. Any series with a little year-to-year momentum drifts, and the eye reads that drift as a wave.
So an honest test needs a fair comparison. Watch what we compare against: for each franchise I measured how strongly one season carries into the next.
One club cleared the five-percent bar. Its supposed cycle had a period of fifty-six years — the entire length of the record we measured.
What is really there: two dials
Killing the cycle leaves something better, because the record does have structure — just not periodic structure. Each franchise is described almost completely by two numbers. The first is its level: the winning percentage it keeps returning to. The second is its persistence: how much of this year's deviation from that level survives into next year. Persistence near zero means every season is a fresh draw. Persistence near one means the club drifts a long way from home before the elastic pulls it back.
| Franchise | Level | Persistence | Temperament |
|---|---|---|---|
| Yankees | .564 | 0.52 | high floor, ordinary elastic |
| Dodgers | .552 | 0.40 | high floor, quick snap-back |
| Red Sox | .537 | 0.04 | good, and with no memory at all |
| Braves | .521 | 0.73 | long slow arcs, in both directions |
| Mets | .497 | 0.50 | average, and stubborn about it |
| Royals | .478 | 0.50 | below average, equally stubborn |
Atlanta is the opposite animal at 0.73. When the Braves go somewhere they stay a while, which is why their history reads as eras rather than jumps.
The rulebook sets the elastic
Now the finding that makes this more than a curiosity. Run the identical calculation on the NFL from 1999 and the whole sport sits somewhere baseball never goes. Baseball's average persistence is 0.42. Football's is 0.16, and several clubs come out negative — actively anti-persistent, where a good year predicts a worse one.
That is not an accident of talent. It is design. Football runs a hard salary cap, a reverse-order draft, huge shared revenue and a seventeen-game season; every one of those is a machine for erasing last year. Baseball has none of them. So football's elastic is short and violent. Baseball's is long enough that a franchise can be recognisably itself for thirty years. When people argue about whether a front office is smart or a market is big, they are arguing about a quantity the league office already largely decided.
There is no clock. There is a level you return to, and a rulebook that decides how fast.
The ProfessorWhat to take home: why this matters outside the ballpark
The cycle is almost always mean reversioni in a costume. Extreme years are partly luck; luck does not repeat; the next year drifts back toward the middle.
A level and an elastic leave us only two things worth arguing about. Can you raise the level? And how long does a good year hold?
Notes & sources
Season records come from the MLB Stats API, 1901–2025; the analysis window is 1970–2025 and covers the twenty-six franchises with at least forty seasons in it. Persistence is the lag-one autocorrelation of a club’s winning percentage around its own mean — a first-order autoregressive coefficient. The cycle test compares each club’s largest periodogram peak against four thousand simulated series matched on that same persistence and variance, which is the fair null: it already contains all the wave-like wandering that momentum alone produces.
One club came in under five percent, at a period equal to the full length of the record, which is a trend rather than a cycle; across twenty-six tests roughly one such result is expected by chance. The NFL comparison uses public regular-season game results, 1999–2025, and the same calculation. Both leagues realigned, expanded and changed schedule length over these windows, which adds noise but cannot manufacture a periodicity that is absent.
The companion primer is Regression to the Mean (Concept No. 3). A dedicated primer on autocorrelation — why a wandering line grows fake cycles, and how to test one honestly — is coming. And a reader-facing toy built on exactly this machinery, which will draw your own fandom trajectory from the teams you pick, is in the works; the arithmetic above is its engine. The archive lives at thesportspage.net.
