The First 15 Metres: Where Split Data Decides Swimming Medals

On 31 July 2026, at Paris La Défense Arena, Pan Zhanle touched the wall in the men's 100m freestyle final in 46.40 seconds. The previous world record of 46.80, set by the same swimmer on 11 February 2026 in Doha, was wiped off the board. The stands erupted, the cameras swept along the lane, and most television viewers kept a single moment: the final touch.

Rewind and pause at the sixth second of that race. The body is still underwater. The legs are still dolphin kicking. The gaps between eight lanes have already been set. Television does not show this segment. The scoreboard does not record it. National teams pay to measure it, hundredth by hundredth, because they know medals are settled there.

I have spent many nights in a small edit room in Miami, replaying 100m freestyle finals at different speeds against the split data returned by electronic timing. What struck me was not the winner. It was that swimming may be the most densely measured sport on earth and simultaneously the most thinly explained one in mainstream coverage.

A hyper-data sport told in three lines

Every lane at a World Aquatics meet is recorded to one hundredth of a second. Sensors on the starting block return reaction time. Sensors on the wall return turn time. Lane-side cameras allow extraction of stroke rate, distance per stroke, and hip trajectory through the underwater phase. In long events such as the 1500m freestyle or the 400m individual medley, the data generated in a few minutes of racing is enough to reconstruct almost the entire architecture of the race.

On broadcast, however, viewers typically receive three pieces of information for a two-minute race: the winning time, the record, and the winner's name.

The gap between the data collected and the data retold is where the craft of swimming analysis exists. I do not argue with emotion; I present a chain of data.

There is a technical reason this retelling is harder than in football. In football, probabilistic models such as xG create an intermediate language audiences can argue with. In swimming, the data needs no model, because it is already measured fact. The problem lies elsewhere: swimming data only holds value when placed beside data from the same era, the same pool, and the same rulebook. Ignore that condition and every historical comparison becomes a puzzle assembled with the wrong pieces.

The sport's history left a clean cut. At the 2026 World Championships in Rome, 43 world records fell in a single meet. The polyurethane suit era ended when the federation imposed a ban from 1 January 2026 and returned to textile. Since then, every record-progression line must be drawn as two separate segments. Anyone merging them into one chart is fooling themselves.

The 15-metre rule and the economics of the underwater phase

World Aquatics rules cap the underwater phase at 15 metres from the wall in freestyle, backstroke, butterfly, and individual medley. In breaststroke, a single dolphin kick is permitted off the start and after each turn before the first arm pull. The 15-metre boundary sounds like a dry technical detail. In practice it is one of the most powerful variables in the sport.

The reason is physics. Underwater, drag falls and the body can hold a higher velocity than it can while swimming on the surface, provided the dolphin kick technique is good enough. A swimmer with a strong underwater phase exits the 15-metre mark at a higher average velocity than the same swimmer does on the surface. In short events such as the 100m, that stretch of time is large enough to settle the ranking before the race truly begins.

This is why leading training centres spend thousands of hours a year on dolphin kick work, ankle flexibility, core strength, and hypoxic tolerance. It is a pure data investment: if the split sheet shows no improvement at the 15-metre mark, the investment is not recorded.

Based on my own experience tracking races and major championships, the difference between a finalist and a non-finalist in the 100m is usually not peak surface speed. It is the stability of the breakout: the variance between underwater phases within a single race, and the ability to hold that quality across heats, semifinal, and final on the same day.

Turns: where tenths of a second get stolen

If the underwater phase is the most celebrated part of swimming data, the turn is the most undervalued. In a 50-metre pool, a 200m event contains only three turns. In a 25-metre pool, the same event contains seven. That discrepancy is not a small detail, because the turn is the action most optimisable through technique and leg strength.

This is also swimming's most common analytical trap: using short-course results to infer long-course performance, or the reverse. A swimmer with outstanding turn mechanics will look far stronger in short course than their overall long-course level warrants. A ranking that blends the two data types is methodologically wrong.

An approximation I still use when briefing editors: imagine a race in which every 25 metres a perfect turn saves three tenths of a second over an average turn. Multiplied seven times, that is more than two seconds. In a 200m event, more than two seconds is the distance between a medal and the heats.

Pacing: what the results board never shows

The results board shows one final time. It does not show how that time was distributed.

Katie Ledecky set the women's 1500m freestyle world record in 15:20.48 on 16 May 2026 in Indianapolis. The value of that swim is not in the result line but in the split structure: she swam the second half faster than the first in a controlled way, while most of her rivals faded sharply over the final 500 metres. Read only the result and you lose all the information about how she won.

Léon Marchand set the men's 400m individual medley world record in 4:02.50 on 23 July 2026 in Fukuoka, breaking a 4:03.84 mark that had stood for nearly two decades. In this event the split structure is far more complex than in any single-stroke race, because the swimmer must convert across four strokes, and each conversion is a cliff edge for efficiency if technique is not maintained. What is worth analysing in Marchand is that he turned the breaststroke leg from a traditional weakness into an attacking segment, then held rhythm in the closing freestyle leg while the body had shifted into lactate accumulation.

I regard the 400m individual medley as the hardest pacing problem in pool swimming. Four strokes, three transitions, four segments with different biology. Anyone distributing effort evenly across all four loses. Anyone emptying the tank in the first two collapses in the freestyle. The optimal curve is asymmetric, and it differs from swimmer to swimmer.

The transfer market of a sport without transfers

Swimming has no transfer window in the football sense. It has an equivalent, and I am spending much of this season tracking it.

There are three flows of personnel. When a head coach leaves a major collegiate programme, the data shows the improvement rate of the remaining squad typically flattens over the following season before recovering or collapsing. Alongside that runs the collegiate transfer portal, where swimmers can change schools once the season ends. The third flow is individual commercial contracts, where a swimmer's value is priced on peak performance, age, and media reach.

Every transfer is a maths problem waiting for a solution.

What catches my attention this season is the data paradox of that market. When a swimmer changes training base, their split sheet travels with them as a file, but the environment that produced the split sheet does not. A breaststroker whose underwater phase is strong thanks to a coach specialising in ankle flexibility does not carry that capability into a new programme. The market prices past performance. The data will answer next season.

The contrarian angle: every copy is wrong on at least one variable

After every major championship, a training template spreads worldwide: dive longer, kick harder, surface later. Thousands of young swimmers do exactly that, and fail.

The reason is a basic analytical error I encounter at least once every season: mistaking correlation for causation. Champion swimmers share one trait, a strong underwater phase. That does not mean the underwater phase produces championships. It is entirely possible that group also possesses height, arm span, innate ankle flexibility, and a rare metabolic profile, with the underwater phase merely a surface marker of a different combination.

Sampling structure is skewed too. We only hold complete split data for swimmers who reach finals. The cohort that spends an entire career training the underwater phase and never reaches a final appears on no data sheet at all. This is classic survivorship error, and swimming is unusually prone to it because the public data pool is thin relative to the internal one.

Meanwhile the era factor is always present. Comparing performances across two decades without filtering for suit rules, pool length, and turn count produces a chart that looks handsome and means nothing. I once built a prediction model for a championship, found an anomalous variable, and spent two weeks verifying the source. The original data was wrong. I did not publish that piece. When an editor says no, I learn to listen to the data, and sometimes listening to the data means accepting that there is nothing yet to say.

One race creates no trend. One championship does not either. But one season, with hundreds of swims under the same rulebook, begins to create signal. Being right too early is also a form of rejection, and in this trade, rejected early is far cheaper than publishing a wrong conclusion that thousands then cite.

The signal for the next cycle

The next stop for elite swimming is the Los Angeles 2028 Olympic Games, and the most valuable data block over the coming two years does not lie with established names. It lies in the split sheets of swimmers aged 16 to 19.

Here is why. If the underwater phase and the turn are where time is created, the gap between a young swimmer with a strong underwater phase and one with only strong surface speed will widen faster than the gap in overall results. That means age-group split data carries higher predictive value than age-group results. Whoever reads that data first holds the information advantage.

Among the noise of the stands, I choose to sit with the spreadsheet. And in swimming, the spreadsheet always starts at the first 15 metres, where nobody applauds, but the medal has already been written.