In the women's 200m freestyle final at the Paris 2026 Olympics, the Omega clock stopped at 1:53.27. The notable detail is not the final number but the 150-200m segment: the swimmer took 29.84 seconds for the last 50 metres, 0.71 seconds slower than the equivalent segment in the morning heat. In a sport where medals are decided by hundredths of a second, 0.71 seconds is the distance between the top of the podium and fourth place. No scoreboard displays the cause. To find it, you have to open the split data and read it lap by lap.

Split data is the set of times for each 50-metre segment of a swim. It differs from the final time in one fundamental way: the final time tells you the result, while the splits tell you how the result was produced. Two swimmers who both finish in 1:55 could have swum two completely different races — one fast out of the blocks then fading, the other conserving energy then surging over the final 100 metres. Looking at the leaderboard, you cannot tell those two apart. Looking at the splits, you see two stories.

I work with 50-metre split data published by World Aquatics and the Omega timing system, cross-referenced against each swimmer's personal record over the past five seasons. The method is not complicated: take each segment time, calculate the gap between the first and last segment, then compare it with that same swimmer's previous swims. I do not care whether someone swims fast or slow on a single occasion. I care whether their pacing structure holds across rounds.

Numbers have no gender, but the people who read them do. A dry split sheet can be read in two ways: as an indictment of weakness over the last 50 metres, or as a map of energy distribution. I choose the second, because the first usually leads to hasty conclusions about a "weak mentality" or "lack of nerve" — things that cannot be measured in milliseconds.

Across the entire Australian team in Paris, I divide swimmers into three groups by pacing structure. The negative-split group — faster in the second half. The positive-split group — slower in the second half. The group that holds an almost perfectly flat pace. At the elite level, the difference between these three groups matters more than a personal best, because it determines who holds form when the pressure rises in a final.

Take Kaylee McKeown. In the 100m backstroke, she swam 58.73 in the heats, 57.99 in the semi-final, and 57.33 in the final. Her pace increased steadily — no sudden jump, no collapse. The notable detail lies in her turn data: her time from touching the wall to surfacing after the turn was about 0.18 seconds shorter than the meet average per turn, and with two turns in a 100m swim, that saves her nearly 0.36 seconds. In a sport where the world record is barely over 56 seconds, 0.36 seconds is an asset.

But the story is not about one athlete. It is about how an entire system distributes energy.

In the women's 200m freestyle, the split sheet shows a repeating pattern: the world-leading group swims the first 50 metres slower than the Australian group but the last 50 metres faster. In other words, Australia's direct rivals — mainly from the United States and China — are choosing a strategy of conserving energy in the first half to unleash in the second. This is data, not speculation. If you take the splits of the top five swimmers over the past three seasons and overlay them, you see a clear trend line: second-half speed is becoming the deciding factor.

Why? Because at this level, first-half speed has almost hit its ceiling. The top female swimmers in the world can all swim the first 50 metres in around 26.5 seconds. The difference between gold and eighth place is not in the first 50 metres — it is in the last 50, where the ability to hold pace as lactate accumulates becomes the real biological boundary.

The key point is this: Australia owns world-leading starting and turning technique, but its pacing structure leans too heavily on first-half speed — a strategy that works in the heats but exposes weakness in the final.

Based on my experience watching matches and championships, I notice something the scoreboard never says: Australia is strong in starting ability and turning technique, but tends to depend too much on first-half speed. This works in the heats and semi-finals, where the goal is simply to qualify. But in the final, when every rival has conserved energy, that strategy exposes its weakness.

The evidence lies in the numbers themselves. If you calculate the average gap between the first 50-metre segment and the last 50-metre segment for the Australian team in the Paris finals, that figure is positive — meaning the last segment is slower. For the American rivals, that figure is nearly zero or slightly negative. This small difference, multiplied across four swims at a championship, creates a large gap in the medal table.

Another metric I track is the conversion rate — the gap between heat time and final time. In theory, a good swimmer should swim the final faster than the heat, since in the heat they usually only need enough to qualify. But the Australian team's Paris data reveals something surprising: in some events, Australian swimmers' heat times were faster than their own final times. This means they went all-out in the heats, or they were affected by psychological factors entering the final.

Take the men's 100m freestyle. One Australian swimmer swam 47.89 in the heats but managed only 48.12 in the final. A gap of 0.23 seconds — not large, but enough to drop two places. Meanwhile, direct rivals swam the final 0.1 to 0.3 seconds faster than the heats. Combined, the gap between the Australian swimmer and a medal is not a gap in physical condition but in the ability to convert form from heat to final.

This is an important finding, and it has strategic implications. If Australia swims the heats faster than necessary, it wastes energy. If it swims the heats slower than its capability, it may not make the final. The art lies in finding the balance — fast enough to qualify, slow enough to conserve. Teams like the United States and China appear to have mastered this problem.

In relay events, the issue is more complex. The 4x100m freestyle relay requires four swimmers to coordinate, and each relay exchange can save or lose hundredths of a second. World Aquatics data shows that the Australian team's reaction times on relay exchanges are among the best in the world — around 0.15 seconds on average, below the 0.20 seconds of many other teams. But a good exchange cannot compensate for a weak leg. And in Paris, some Australian legs were significantly slower than those swimmers' own personal bests.

One more factor that splits do not display directly but that can be inferred: underwater technique. After each turn and after the start, a swimmer can travel underwater before surfacing. In backstroke and butterfly, the effective underwater distance is usually longer. Data from recent meets shows that top swimmers travel about 10 to 15 metres underwater after the start, while the average group manages only 7 to 10 metres. This difference, multiplied by underwater speed being higher than surface speed, creates a cumulative advantage.

There is a temptation any analyst must guard against: turning correlation into causation. We see a swimmer go slower in the last segment and lose, then conclude that a slow last segment caused the defeat. But if you take ten years of data, you will find countless cases of swimmers who went slower in the last segment and still won, and countless cases of swimmers who went faster and still lost. Correlation only tells us two things happen together, not which causes which.

Kazan is the day I learned that a 99% probability can still die on the betting table. In 2026, I analysed the Germany versus South Korea match and pointed out that Germany had 74% possession but only 11 passes into the box. I called it the arrogance of the rich refusing to press. German fans attacked me. A week later, FIFA published data confirming every number. But the bigger lesson lay elsewhere: correct data does not guarantee a correct conclusion. A small sample, an unusual context, an omitted variable — all can overturn a beautiful model.

In swimming, this problem is even more serious because the sample size is small. A swimmer only races their main event a few times a year at the highest level. Three swims are not enough to conclude a trend. If I say a swimmer tends to fade in the last segment, I must ask: based on how many swims, in how many contexts, with how many variables controlled? Many conclusions I read online rest on just one or two swims — and that is fortune-telling, not analysis.

I do not believe in emotion. I believe in a data series longer than your emotion. But I also know that a long data series can still lead to a wrong conclusion if the person reading it does not know how to ask the right questions.

The limits of data

There are things splits cannot measure. They cannot measure the fear of the water in a final. They cannot measure a swimmer waking at four in the morning from anxiety. They cannot measure the pressure a nation places on the shoulders of a twenty-year-old girl. They cannot measure the feeling of water half a degree colder than usual in the pool, or a coach changing the training plan two weeks before a meet.

I say this not to deny the value of data but to put it in its proper place. Data is a map, not the territory. It tells us the terrain but not the feeling of walking it. And in a sport where the gap between medals is measured in hundredths of a second, the things that cannot be measured are often the things that decide.

What is worth watching in the next cycle is not who swims fastest in the heats. What is worth watching is who holds their pacing structure when the pressure rises — and whether Australia adjusts its energy-distribution strategy to swim the second half faster. If it does, that 0.71-second gap can change sign. If not, we will again see beautiful split sheets in the heats and bitter endings in the finals.

And perhaps the real question is not how to swim faster, but how to understand what is truly happening beneath the surface. Because data only answers the questions we know how to ask.