In a women's 100m freestyle final, the gap between gold and fourth place is often under a tenth of a second. But there is another number rarely mentioned on broadcast: the underwater distance after the start of those four athletes differs by nearly two meters. Two meters underwater, at two meters per second, is a second. A second, in a women's lane, is an entire career. I spent the past season breaking down every underwater segment of the women's semifinals and finals, and what I found forced me to rewrite how I read a race.

Data has no gender. But the people who read it do, and those readers — coaching staff, bookmakers, television audiences — usually look only at the final number on the board. They see 52.04 and call it speed. I look at the 0-15m split and call it the decision.

Context: why this season deserves to be measured from scratch

Over the last three cycles, women's swimming has undergone a quiet but systemic shift: the center of gravity of the race has moved from the middle of the pool to both ends. Starts have been refined, turns optimized, and especially the underwater segment — where athletes perform dolphin-kick sequences before surfacing — has become a discipline of its own.

I collected data from Australian national meets and several international rounds during the annual season, using split data published by organizers alongside high-frame-rate footage to re-measure segments that lack available numbers. My method is always the same: state the research question, define the sample size, then conclude. This year's question was simple — is peak speed within a race still the best predictive variable for final ranking?

The answer, after calculation, is no. And that is why I am writing this.

Core: the data evidence chain

I divided each women's race into five segments: start and underwater (0-15m), surfacing and acceleration (15-25m), maintenance (25-50m, depending on distance), turn, and finish. I then cross-referenced each segment against the final result.

The first thing I found: in my recorded sample, the correlation between underwater distance in the first 15m and final ranking was stronger than the correlation between measured peak speed in the middle of the pool and final ranking. In other words, the athlete who surfaces later but keeps momentum often finishes higher than the one who surfaces early with higher peak speed in the middle segment.

The specific number that caught my attention: among the women's 100m freestyle finalists, those with a total underwater distance above 12 meters across two starts and turns made up most of the leading group. The early-surfacing group, under 9 meters, typically lost position in the 25-40m segment even when their acceleration split was not bad.

But here is where I must be careful. Correlation is not causation. Swimming more underwater does not automatically produce a faster athlete. The reverse may hold: the strongest athletes, with the best kick and fitness base, are the ones able to sustain a long kick sequence without losing speed. Underwater distance then is only a marker of capability, not a cause of performance.

To separate these two possibilities, I examined the average speed within the underwater segment itself. If an athlete swims 14 meters underwater but slows markedly in the final 4 meters, the value of that long segment is nullified. If another athlete swims only 10 meters underwater but holds speed nearly constant until surfacing, that athlete usually transfers momentum better into the freestyle segment.

In other words, the right variable is not underwater distance, but the quality of that distance. This is the kind of distinction a raw data table, recording only total meters, erases. And this is precisely the blind spot of most analyses I read during the season.

I once wrote that the heat map has become the new fortune-telling of sports. The lane is the same. People color the fast segment red, the slow segment blue, then pronounce on an athlete without knowing what decision that athlete made at meter twelve — where no clock records anything, only a mind calculating in water.

Emotion is also data

There is one thing I must concede, even though it does not fit neatly into a spreadsheet. The decision to surface early or late at meter fifteen is not only a physics problem. It is a decision under pressure. An athlete in lane 5, leading, hearing the crowd behind, sensing the water beside her advancing — that person surfaces early not because of poor technique, but because of fear.

Fear is a variable. It has no unit. No sensor measures it, and no bookmaker prices it directly. I do not trust emotion. I trust a data series longer than your emotion. But I also know that even a long data series can miss the single thing that produced it.

In Kazan, I learned that a 99% probability can still die on the betting table. That lesson did not teach me to stop using data. It taught me to state my limits clearly. In the lane, the limit is this: I can measure distance, I can measure speed, but I cannot measure the fear in the three seconds before surfacing.

Contrarian: when beautiful data is a warning sign

There is a paradox I found when cross-referencing data across multiple rounds. The athlete with the most beautiful underwater segment in the heats is not always the one who wins the final. The group that surfaces with the highest peak speed in the heats tends to reproduce that speed in the semifinals, but loses it in the final. Meanwhile, the group with a stable but unspectacular underwater segment often climbs the rankings as pressure increases.

The most plausible explanation, based on my data, is this: beautiful heat data may reflect a distribution of effort that is not yet sustainable. An athlete who over-invests in the start to make an impression will pay for it in the final stretch of the final, where equally matched rivals have adjusted. Stability, rather than peak, is the truly predictive variable in the knock-out rounds of swimming.

This runs counter to media intuition, which always favors the biggest number. The biggest number sells news. But the stable number buys medals.

I must also concede a major limitation: my sample size is insufficient to fully eliminate the influence of pool size, water temperature, or a dense daily schedule. An athlete competing in three events over two days has a completely different start segment than one competing in a single event. No model of mine currently isolates that variable.

Takeaway: three zones of a lane

After this season, I redraw the map of reading a women's race into three zones. The first is the zone of affirmable data: underwater distance, average speed per segment, rate of deceleration — all measurable and predictive. The second is the zone of ambiguous data: the correlation between underwater distance and ranking, where causation is entangled and more data is needed to separate. The third is the zone that must rely on intuition: the moments in the water, before surfacing, where decisions are made and leave no numbers.

What I carry into the next cycle is not a formula, but a question. If peak speed is no longer the best predictive variable, then what are the analytics industry, the bookmakers, and the broadcasters mispricing in women's swimming? The answer, I believe, lies in the meters no one rewinds to watch — and that is exactly where I will spend next season counting again, slower, more carefully, and more humble about what I cannot yet measure.

Limits of the data: I cannot measure morale, day-of form, the influence of the starting official, or water quality. My sample is small and not representative of the entire global women's swimming system. The correlations above are not presented as causal relationships.