On Tuesday evening I opened the results file of a short-course meet in Queensland and found the 50m split column empty. Not missing a few rows. Empty throughout. Fourteen events, three hundred and twenty-six swims, and not a single figure in the split-time column. The colleague beside me offered one line: “Just write it by feel, nobody reads the tables anyway.” I closed the laptop. Across eleven years of doing swim analysis for the Australian market, I have received a file like that four times, and four times I have had to choose between inventing a story and admitting I had nothing to say.

Numbers have no gender. The people who decide to fill the gaps in a spreadsheet do.

Swimming is the most densely measured sport on the Olympic programme, and also the most misunderstood where data is concerned. A 50-metre lane generates hundreds of measurement points: reaction time off the blocks, entry time, underwater distance, dolphin-kick count, turn time, stroke rate, distance per stroke. World Aquatics publishes official results with 50m splits for most major meets, and since late 2026, when the organisation rebranded from FINA, it has tightened its publication standards. But the distance between what is published and what is actually measured remains as wide as the gap between a medal and a 5am training session.

My job is to bridge that gap for bookmakers and Australian broadcasters. A swim at state level may carry six measurement points. A swim in an Olympic final may carry forty. The same athlete, the same event, two datasets with wildly different resolution.

Two things need separating here, and most sports writing conflates them. A sparse dataset has few measurement points, and from it you can still infer something with low confidence. A null dataset has no measurement points at all, and from it you can infer nothing. The first lets you write with an error warning attached. The second only lets you write with imagination.

I once sat in a press box in Brisbane, published an against-the-grain call built on xG of 2.4 against 0.6, was laughed at by a commentator, and watched the call land. I have also stood in the stands at Kazan and watched Germany collapse against South Korea with 74 percent possession, and learned there that a 99 percent probability can still die on the betting board. Both memories taught the same lesson: the danger is not bad data, it is a story built on an empty table.

Pan Zhanle swam 46.40 seconds in the men's 100m freestyle final at Paris 2026, breaking the world record he had set four days earlier on the lead-off leg of the 4x100m relay. The figure alone says nothing about how it came about. The 22.28-second opening 50m is the anchor. In the history of this event, very few swims under 47 seconds have a first half under 22.50. The performance came from the back half, not from raw speed up front. That is the kind of information a split table provides — and the kind an empty table removes.

At the same Olympics, Bobby Finke swam 14 minutes 30.67 seconds in the men's 1500m freestyle, breaking a world record that had stood for twelve years. In an event that long, every swimmer crosses the 50-metre mark thirty times. Thirty measurement points per swim is a dataset thick enough to separate a badly paced race from one that accelerated at exactly the right moment. Without those thirty points, any analysis of Finke is reduced to describing a feeling.

Katie Ledecky won the 800m freestyle at Paris 2026, her fourth consecutive Olympic title in the event, taking her Olympic gold tally to nine. More interesting to anyone working with data is how narrow the variance in her times has been over her specialist distance. A sequence that long lets you separate two things the eye cannot: baseline capacity and day-form.

Some things are worse than an empty table: a table used badly. The first metric I check when assessing a swim is not average speed but split structure. Two swimmers with identical finishing times can have completely different split structures: one goes fast early and fades, one goes slow and detonates late. Same finishing time, opposite predictive value for the next swim. Bookmakers look at time. Modellers look at structure. That is why heat maps and average-speed charts have become the new astrology of sports analysis: they compress several layers of information into one number, then sell that number as a conclusion.

In swimming, the most neglected part is the part underwater. From the start, a swimmer may travel up to 15 metres submerged before the first breakout. In butterfly and freestyle, that submerged segment is where the highest speed of the entire swim occurs. Yet most public data contains no column recording actual underwater distance. Analysts have to reconstruct it from video, and that work is slow enough that almost nobody does it properly.

Put another way, the most expensive thing in swim analysis today is not the model. The most expensive thing is data that was never collected.

The counter-intuitive point sits here: a null result is still a result. When I received that empty file, I had no information about the swimmers' technique, but I had information about the organisers. A meet that records no split times either has a timing system not configured to export them, or has nobody paying for the export. Both possibilities say something about how the organisers value the competition they run.

The second counter-intuitive point: the biggest temptation comes not from too little data but from too much. At a meet like the Olympics you have splits, stroke rates, turn times, touchpad sensor data. That abundance creates the illusion that every question has an answer. Some of the most important variables appear in no table at all: how much an athlete's shoulder hurt after yesterday's session, whether she slept four hours from nerves, whether the coach cut the volume on exactly the right morning. Those decide a swim, and no sensor measures them.

I do not trust emotion. I trust a sequence of numbers longer than your emotion. But however long the sequence, there is always a stretch you cannot read, and the only honest response is to mark that stretch clearly.

Limits of the data: this piece does not include the split table from the Queensland meet mentioned at the start, so its technical section is deliberately left blank. The figures for Pan Zhanle, Bobby Finke and Katie Ledecky come from official World Aquatics results at Paris 2026. The underwater-distance section rests on video observation, with an error margin of up to one metre. Competition psychology and injury status appear in no public dataset, and I do not speculate on them.

Brisbane 2032 lies ahead. The question I set myself for the next cycle is not how many more gold medals Australia will win in the pool, but how many data columns its national championships will export in the results file. If six months from now I open an Australian national results file and find the split column full, that will be the signal I track more closely than the medal table.