Two hundredths

The two best years in the history of swimming belong to the same man, back to back — and the title that made each of them came down to one hundredth of a second. Also: the number everyone quotes about him is fake.

The year everybody remembers is 2008 — eight gold medals in Beijing. It is not his best year. By the measure below, the best year anyone has ever had in this sport is Michael Phelps in 2007, and the second best is Michael Phelps in 2008.

The measure is this: in a single calendar year, how many separate individual events did you finish ranked number one in the world in? Not a medal. Not a final. The fastest time on earth that year.

Now look at the last column of each table, and find the smallest number.

In 2007, the title that takes him from 5 to 6 is the 100m Freestyle, won by 0.01 seconds — and second place that year was shared, two swimmers on the same hundredth. In 2008, the title that takes him from 4 to 5 is the 100m Butterfly, won by 0.01 seconds over CAVIC Milorad.

Erase two hundredths of a second from the history of this sport and the greatest run anyone has ever had becomes 5 and 4 — and 4 is a number 9 other athlete-years have also reached.

Why there is no “one in a billion” in this piece

The usual way to write this story is to compute how many standard deviations the subject sits above the field, then convert that into a probability. It produces a huge, quotable number. It is also wrong, and this dataset shows exactly why.

Here is the same query — Men's 200m Butterfly, season world ranking — run for two different years:

The sport did not get 799 times deeper. That is digitisation, not participation. Any z-score computed against “the field” is therefore mostly a measurement of how much of an era somebody has typed into a database — and a normal curve extrapolated five or more σ into the tail of a bounded, selection-thinned distribution is wrong by orders of magnitude in a direction nobody can bound.

So this analysis never computes one. Rarity here is a count of things that happened. Across 1985–2025, both sexes, all 17 individual long-course events, there were 865 athlete-years in which somebody finished a year ranked #1 in the world in at least one event. This is how they are distributed:

Two is not half of four

The obvious objection is that the count is arbitrary — that being number one in five events just means entering five events. It does not, and the file says why. Look at who else managed multiple world #1s in one year:

They are pairs like the 100 and the 200 backstroke, or the 50 and the 100 freestyle — one stroke swum twice, by an athlete whose training, physiology and race craft transfer almost perfectly between the two. That is not a marginal pattern: 82% of all two-title athlete-years in the file are a single stroke family.

Which is what makes the ceiling interesting. Nine athlete-years have reached 4, and they are not all the same shape:

Phelps’s two record years span 3 stroke families — IM, fly, free — across a range from a fifty-second sprint to a four-minute medley. Spanning three strokes is not unique; Lochte, Otto, Marchand and McIntosh all did it. Doing it while holding more than 4 titles is, and that is the claim the analysis actually gates.

The gates

Six checks run before the dataset is allowed to be written. They exist so that a claim in this article cannot quietly stop being true.

• G0 — known answer. Phelps’s 2008 must come back as world #1 in exactly the five events the historical record says, before anything is derived from the parser.
• G1 — coverage floor. An event-year counts only where at least 25 swimmers are ranked, so “#1 in the world” is never an artefact of a six-row table.
• G2 — supersuit independence. 2008–09 was the polyurethane era and the headline year sits inside it. Delete both years entirely and the record must still stand. It does.
• G3 — rarity is a count. The reported rarity must equal an observed number of athlete-years, asserted structurally so no later edit can swap in a fitted tail.
• G4 — threshold robustness. Re-run at a top three cutoff instead of #1; the subject must still hold the maximum. A result that exists only at one cutoff is a cutoff.
• G5 — breadth. The subject’s events must span strictly more stroke families than any other athlete-year in the file.

What this does not say

It does not say Phelps would beat a swimmer from another decade; nothing here compares across eras, and the unit is always “ranked #1 among his own contemporaries that calendar year”. It does not pool the men’s and women’s rankings into a single comparison. And a season world #1 is the fastest time swum in a year, which is not the same thing as beating everybody in a final — it happens that most of these were Olympic finals, which is stated rather than assumed.

The archive is also not the sport. Coverage before the mid-nineties is thin even for major events, and the coverage floor drops those event-years rather than counting a shallow one. Anyone who was genuinely the fastest in the world in a year the archive barely holds is missing from this file, not demoted.