Swimming Splits: How a Data Gap Distorts an Entire Conclusion

Swimming Splits: How a Data Gap Distorts an Entire Conclusion

At a national-level swim meet in June, I sat in the fourth row above the starting blocks with a stopwatch in one hand and a sheet of paper divided into eight boxes in the other. The swimmer in lane 4 touched the wall and the electronic scoreboard reported a personal best: 1.2 seconds faster than her own time three months earlier. The stands applauded. My sheet recorded a very different decomposition: 0.38 seconds came from the first 15 metres underwater, 0.21 seconds came from the third turn, and the arm stroke between turns barely moved at all.

Read the scoreboard and you conclude she swam faster. Read the splits and you conclude she surfaced better and turned better, while her clean swimming speed stood still. Those two conclusions lead to two different training programmes, and only one of them matches the pool.

Swimming is the most precisely measured of all time-based sports, but only at the top level. At the World Aquatics Championships and the Olympic Games, World Aquatics publishes result files with 50-metre splits and start reaction times. At every other level — national championships, junior meets, school cups, where most swimmers actually improve — the only thing recorded is the final time on the board.

That gap decides how the story gets told. Football has xG to force writers to separate process from outcome. Swimming has splits, but only when an organiser pays for a touchpad system that logs data. The consequence is that most swimming writing, including specialist writing, is really just the scoreboard retold with adjectives.

Two technical traps sit inside that gap, and both generate systematic error. Pool size is the first. A 200-metre race has seven turns in a 25-metre pool and only three in a 50-metre pool. If the underwater phase and the turns account for roughly a third of any speed advantage, a swimmer with a strong kick will look dominant in short course and ordinary again in long course.

The other trap is the suit era. Between 2026 and 2026, when polyurethane suits were legal, a run of world records was set that still stands on the books today. Paul Biedermann set the men's 200-metre and 400-metre freestyle world records in Rome in 2026. Federica Pellegrini swam 1:52.98 in the women's 200-metre freestyle, also in Rome in 2026. Any all-time list that mixes those marks with fabric-suit-era results is comparing two different sports.

With full splits available, a race divides into four independently measurable blocks.

The cheapest block is start reaction time. At elite level it sits between 0.60 and 0.72 seconds and is fairly stable for any individual, which makes it barely a training variable. A swimmer shaving 0.05 seconds off reaction time after a year of dedicated work is a rare case. When an article attributes a breakthrough to a "better start", I check this field first, and most checks explain nothing.

One step more expensive is the underwater phase, made up of the push-off after the start and the kicks after each turn. The rules require the head to break the surface before the 15-metre mark, turning that distance into a race of its own. This is where technique yields the largest advantage, and also where public data is thinnest, because most organisers only record splits at the 50-metre wall.

The hardest block to measure is stroke rate and distance per stroke. Swimming speed is the product of the two, so raising stroke rate without holding distance per stroke merely trades one unit for another. In distance freestyle, rate typically sits around 40 to 50 cycles per minute; in sprint events it is far higher, at the cost of shorter distance per stroke. Without wearable measurement, every analysis of this block is inference from video.

The most neglected block is the wall touch. In short course, where the number of turns doubles, one weak touch can erase the advantage accumulated across the previous 150 metres.

Seen across all four blocks, a principle emerges clearly: an improvement in time should only be called technical progress when it appears in the clean-swimming segment, not in the start, the underwater phase or the turns. The first three blocks depend heavily on neuromuscular state on the day, on pool depth, on the quality of the starting wedge. The clean-swimming segment is what accumulates across thousands of hours and is least disturbed by outside conditions.

I hand-timed the splits at that meet, and the error of that method is roughly 0.1 seconds per mark. Any gap below 0.15 seconds on my sheet falls into a zone where no conclusion is possible. When an editor says no, I learn to listen to the data. What the data says here is that most conclusions about swimming below the elite tier are built on a foundation whose error is larger than the effect being measured.

Having splits is still not enough. A swimmer who turns 0.2 seconds faster at one meet may simply have reduced training volume two weeks earlier, and that taper also improved the other three blocks at the same time. Blaming the turn is a correlation dressed up as causation.

The larger problem is an empty data input. No splits, no reaction time, no 15-metre marks, not even a publication date for the results — at that point any deep analysis becomes fiction with a tidy interface. The only honest handling is to stop and say the data is insufficient, rather than filling the hole with plausible-sounding guesswork.

Swimming adds one more layer. The credibility of the sport rests on an anti-doping system with traceable records. A story built purely from a final time, with no process data and no testing history, inflates an ordinary result into an event larger than the data permits. I do not argue with emotion; I present a chain of data.

The signal worth following is not a new record. It is whether national federations add a split column to their electronic result files, and whether organisers publish 15-metre marks as a standard data field. Among the noise of the stands, I choose to sit with the spreadsheet — and one more column would open a different solution to the whole problem.