The Race Without Data: The Three-Source Verification Discipline in Swimming Analysis

On 31 July 2026, at La Défense Arena in Paris, the scoreboard showed one number: 46.40 seconds. Pan Zhanle broke the men's 100m freestyle world record. The next morning, many reports repeated that number, then stopped. Open the split sheet, and the story turns. First 50m: 22.28 seconds. Second 50m: 24.12 seconds. Pan swam the fastest opening lap in history, then finished at the speed of an ordinary finalist. The 46.40 figure is not wrong. It simply does not say everything. After more than twenty years reading swimming data, I have learned one thing: a bad data sheet still has room to be saved. A blank data sheet filled with guesswork does not. An elite swim does not produce one number. It produces hundreds. Omega's timing system — Olympic official partner since 2026 — records touchpad times, reaction times and splits at every 25m or 50m. Add video data, swim-speed data, stroke rate and distance per stroke. A 200m medley final can emit several hundred individual data points. What reaches spectators is usually one line: final time and ranking. In Vietnam, that gap is wider. When Nguyễn Huy Hoàng swims the 1500m freestyle, or Trần Hưng Nguyên reaches a 200m medley final, detailed split sheets are rarely published in full for domestic media. Analysts must reconstruct the race structure from broadcast footage, hand timers and coaches' notes. Three different sources, one swim — and they do not always agree. That is why I set my own rule: no number enters an article before it has passed at least three independent cross-checks. This profession does not allow guessing. With enough splits, a 100m freestyle swim splits into four fundamentally different segments. The first 15m is start and underwater data. To 50m is acceleration data. From 50m to 85m is maintenance data. The last 15m is touch data. Pan Zhanle in 2026 won in the second segment, not the fourth. That reading completely changes conclusions about how he trains. It suggests his strength lies in generating enormous momentum right after surfacing, not in enduring pain at the end of the race. By the same logic, reading only the final time of a 1500m swimmer tells us nothing about whether he won by kicking early or by holding rhythm and unleashing a 200m sprint. Those two kinds of winning demand two entirely different training plans. The split sheet is the only thing that tells them apart. Swimming, therefore, is a sport where structure matters more than the final result. A sheet with only a final time is like a contract that lists the total value but omits every clause. You know the price, but not why it costs that much. And here the swimming analysis trade meets the problem of every data industry: when the data pipeline breaks, you are not allowed to invent. You are allowed to say "not yet readable". In 2026, when international swimming nearly froze, I had a chance to review how I built models. An empty dataset is usually handled in three ways. First: fill it with interpolation. Second: ignore it and conclude as usual. Third: mark the unreadable zone and fence it off. The first two lead to the same outcome: a model that is clean, tidy and wrong. I once treated models as scripture. Now they are only a compass — but without one, we are lost. A good swimming model must declare where it is blind. If a model predicts a 200m medley time without split data at 150m, the error margin in that final turn must be clearly marked, not hidden inside a pretty decimal. In my own analyses, I always keep a section called the "unreadable zone". It lists the things I lack sources to assert: an unconfirmed shoulder injury, a training-plan change with no interview, a fitness test with only one source. That section is ugly. It makes the writing look indecisive. But it is honest, and over time it is the thing that preserves credibility. There is a temptation every analyst has met: see a world record and attribute it to a training method. Pan breaks a record, and immediately there are tributes to the training model of an entire sporting nation. In reality, a record is the result of one individual on one day, under one set of equipment rules, in one pool. Correlation is not causation. A swimmer being fast does not prove his training plan is right; it only proves he is fast. To talk about a training plan, you need data across many years, many athletes and many meets — including the athletes who failed under that same plan. Sports data calls it a control group. Very few articles have one. Numbers do not lie, but people always find ways to lie with numbers. The most common lie is to pick one striking figure and attach to it a cause with no evidence. In swimming, that lie often appears under the cover of "technical analysis". A slow-motion clip cut at the prettiest frame, one perfect surfacing, and a conclusion is drawn for an entire long-term training plan. No splits, no control, no sample. Just one beautiful frame. When the pool is empty, every model collapses. I rebuild from the half-burned data. For Vietnamese swimming, that half-burned data still holds many usable pieces. We have competition times from domestic meets, youth athlete records season by season, enough to reconstruct the improvement curve of a swimmer over three or four years. Those pieces are enough to ask the right questions, even when not enough to answer them. The work for the next cycle is not to publish another medal prediction. It is to build a minimum data standard for every recorded swim: are there splits, is there a reaction time, which source confirms it. Only when we know what we are missing do we know what we are saying. Reputation is only a name. What remains is always how you read the race — even when the data sheet is blank.