Falling records
A Socratic walk-through of falling records — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do athletic records keep improving when the human body has not changed?
A gold medal time from 1924 would not qualify a swimmer for a decent club meet today. The obvious reading is that humans have improved. But a century is nothing in evolutionary terms — the genome of the species has not meaningfully moved since 1924 — so whatever improved, it was not the body's design. What was it?
Reasoning it through
REASONING #Begin with what a record actually is, because the phrase hides the trick. A world record is not the performance of a typical human. It is the single most extreme performance ever produced by anyone who competed. It is a maximum drawn from a sample.
That reframing does a lot of work immediately. There are two entirely different ways to raise a maximum. You can shift the whole distribution upward — make everyone better. Or you can take more samples from the same distribution, because the more people you draw, the further into the tail you are likely to reach. Which of those has happened more dramatically since 1924?
Consider who was eligible to appear in the sample. Early modern athletics drew from a thin slice: men, in a handful of wealthy countries, with enough private means to train while not being paid, in the events their schools happened to run. Since then the sample has expanded on nearly every axis at once — most of the world's countries, women's events across the programme, professional pay that lets an athlete train as a job rather than an evening hobby, and deliberate talent-identification systems that go looking for the rare physiology instead of waiting for it to wander in. The pool did not grow by a percentage. It grew by orders of magnitude, and the tail of a much larger sample sits much further out.
Now add the distribution shift, which is real too. A 1924 athlete trained on folk wisdom. Since then: interval training and periodised programmes, measured physiology, nutrition that is planned rather than assumed, sports medicine that returns an injured athlete instead of ending them, and coaching knowledge that accumulates and transmits rather than dying with each generation.
Then equipment and surfaces, which are easy to underrate. Cinder tracks gave way to synthetic ones. Bamboo and steel vaulting poles gave way to fibreglass, which changed the event so completely that the old marks became irrelevant. Pools got deeper with better lane dividers to kill turbulence. Distance-running shoes acquired stiff carbon plates and resilient foams.
But now a harder question. If all of this were true, would records fall forever? And is the fact that they keep falling itself slightly misleading?
The analogy
THE ANALOGY #Think of a record as the highest number ever rolled in a very long game of dice. Roll more dice and the highest number rises, even though no individual die has changed — that is the growing pool. Weight the dice slightly and it rises further — that is training, nutrition and equipment. But no amount of either takes you past the highest face on the die, so the sequence of new highs comes further and further apart as you approach it.
dice have a hard ceiling that is known in advance, and human limits are neither hard nor known — the ceiling is a soft biomechanical and physiological boundary we can only infer from how the curve is flattening. And unlike a die, the athlete really is being improved, so the distribution and the sample size are changing at the same time.
Clarifying the model
THE MODEL #Three things sharpen this, and one of them is uncomfortable.
First, the ratchet. A record can only be broken, never un-broken. So a chart of records rises monotonically by construction, and would keep rising even in an era when the average elite athlete was getting no better at all — every unusually good day is kept forever, and every ordinary day is discarded. Some of the impression of relentless progress is the recording rule, not the athletes.
Second, deceleration. In many long-established events the improvements have become small and slow, and the curve is visibly bending toward an asymptote. Several men's track records have stood for decades. That is what the dice analogy predicts, and it is the strongest sign that we are near limits rather than moving steadily away from them.
Third, and this is the uncomfortable part: some records are suspect. Several standing women's track marks were set in the 1980s, under testing regimes that were weak where they existed at all, and have not been approached since — which makes the historical curve in those events unreadable as a record of legitimate progress.
The cleanest evidence for the technology channel is a natural experiment. When full-body polyurethane swimsuits were permitted in 2008 and 2009, world records fell in absurd numbers, dozens at a single championship. The suits were then banned from 2010, and many of those records stood untouched for the better part of a decade while athletes who were plainly faster swimmers could not reach them. Remove the equipment, and the progress reverses. That is as close to a controlled test of "was it the athlete or the kit" as sport is ever likely to offer.
A picture of it
THE PICTURE #How to readRead left to right as eras rather than exact dates, with each entry giving the change first and its visible consequence second. The first four entries are all about the sample — who was allowed in, who could afford to train, and who went looking for talent — while the later ones are increasingly about equipment. The 2008 to 2010 pair is the important one: it is the only place in the sequence where a cause was added and then deliberately removed, which is why it tells you more than the rest of the line put together.
What became clearer
WHAT CLEARED #Records fall because the sample got vastly larger, the training got better, and the equipment kept changing — not because the body did. The dice framing explains both halves of the pattern at once: more draws and slightly weighted dice push the maximum out, but the ceiling does not move, so the increments shrink as we approach it. And the ratchet rule, which keeps every good day and discards every ordinary one, quietly flatters the whole curve.
Where to go next
ONWARD #- How statisticians estimate an event's ultimate limit from the shape of its record progression.
- Why some events (marathon, pole vault) show step changes at equipment transitions while others improve smoothly.
Key terms
TERMS #| Term | What it means |
|---|---|
| Extreme value | the maximum of a sample, which grows with sample size even when the underlying distribution is fixed. |
| Talent identification | systematic screening of a population for the physiology an event rewards, rather than waiting for self-selection. |
| Asymptote | a limit a curve approaches with ever-smaller increments without reaching it. |
| Natural experiment | a real-world change that isolates one variable, as the 2010 swimsuit ban did for equipment. |
Every term the collection defines is gathered in the glossary.