The Blank in Swimming Data: Why an Analyst Must Know When to Stop

Tuesday morning, I was sitting in the third row of a 50-metre pool, a split sheet for a group of junior swimmers in my hand. An acquaintance stopped by, pointed at lane four and asked: "Can that kid break the SEA Games record?"

I opened the dataset. The time column was empty. The stroke-count column was empty. The final-50-metre speed column was empty. Empty from the first cell to the last. No athlete name, no event, no competition date, no meet. For about three seconds I felt the pressure of the job very clearly: say something that sounds convincing.

I said: "Not enough information to assess."

The acquaintance nodded, looked disappointed and walked off. That answer does not sell a story. But it is the only fence standing between analysis and guesswork.

Vietnamese swimming has results, but it lacks data

In Vietnam, the elite swimming data the public can reach is thin. Final results from national championships, the SEA Games and the Asian Games usually offer finish times and little else. The things that build a technical story — 50-metre splits, stroke rate, distance per stroke, reaction time off the blocks, underwater time after each turn, speed over the final 15 metres — are almost never published.

A few training centres hold that data. It lives in internal files, in coaches' spreadsheets, on personal laptops, and it rarely leaves the training hall.

So every time a swimmer performs well, the question of "why" gets answered with something else. With a coach's feel. With a commentator's memory. With dressing-room rumour. All three sources have value, but they belong to a category of data that cannot be cross-checked.

I work as a data consultant. My job is to reconstruct a race, a training cycle, a substitution decision from numbers that can be traced. When the input is blank, the whole system collapses to a single point: there is nothing to say.

Every shock has its own probability. We call it a shock only when we have not yet checked the table.

Nine analytical layers, nine traps

The process I use splits a deep analysis into nine layers: technique; performance and data; competition system and selection mechanism; the global map of the sport; rules and anti-doping governance; athlete career and team system; risk profile; public narrative and expectations; and the industry ripple.

When the upstream extraction layer returns a blank, those nine layers turn into nine traps. Each one has a highly plausible answer ready, and each of those answers is wrong.

Take the technique layer. Without splits you cannot say anything about closing speed over the final 100 metres. Without stroke count you cannot tell whether a swimmer is going long or short. A coach can see a great deal with the naked eye, but the human eye cannot measure a three per cent drop in speed over the last 25 metres of the third lap.

Take the performance layer. Without a finish time there is no placing, no comparison with the national record, no sense of where an athlete sits between an A cut and a B cut.

Take the competition layer. Without the name of the meet you do not know whether this is an Olympic year, an adjustment year or a build-up year. Results read completely differently in each.

Take the career layer. Without a name, an age or an injury history you cannot discuss the puberty barrier — a familiar pattern in Southeast Asian swimming, where many female athletes shine between 13 and 15 and then leave the lane when their bodies change. Staying at the top through that threshold is the exception. Nguyen Thi Anh Vien is a rare example of a career that lasted across several cycles, with eight gold medals at the 2026 SEA Games in Singapore according to the official results of the organising committee.

Take the rules and anti-doping layer. This is the most dangerous one. Raising doping without a single document behind it puts a human being on trial in the media. A doping suspicion has to be classified precisely: a confirmed violation, a contamination dispute, a procedural error, or simply an online allegation. No document, no layer.

The other four layers behave the same way. There is no anchor point for the analysis.

A tactical era dies when nobody reads its data table any more.

The blank is itself a data point

What most viewers never notice: a system returning a blank is data about the data system itself.

When the first extraction layer comes back empty while the domain label still appears, the most likely explanation is an extraction fault — bad encoding, a truncated document, a misaligned field map. A real article was dropped somewhere along the way. If every field is blank, including the label, then the input source was genuinely empty.

Those two scenarios lead to two different actions. One is a technical fix. The other is to stop and go find the source.

The one thing that must never be done: fill the blank with a story that sounds reasonable.

The bigger trap: correlation dressed as causation

Sports data has an old scar. When two series move together, people rush to conclude that one causes the other.

A swimmer improves during the same period in which the coach changes. A relay squad breaks through after a new programme. A centre increases training sessions and the medal count rises. All three may be true. But before saying "X leads to Y", I am obliged to name the physical or behavioural mechanism connecting the two variables. If I cannot name it, the honest phrasing is "there is an association".

In swimming, the mechanism is usually very concrete. Drag falls when the body line is straighter in the water. Stroke rate rises when the shoulder musculature is stronger. Underwater time lengthens when the push-off technique and lung capacity improve. With a mechanism, there is causation.

A contract is not a signature. It is a hypothesis with a name on it.

Over more than twenty years in the third row of swimming pools, I have finished a beautiful model many times and then torn it up after thirty minutes with a coach. A data analyst always risks concluding at a rhythm that is out of step with the reality of the locker room.

The variance that cannot be measured

There is a part of Vietnamese swimming that data never touches.

The stands. Emotion. A family's expectations. The pressure of a single international slot that comes once every four years. A 17-year-old stepping onto the blocks while an entire province watches on a phone screen.

My model treats crowd noise as a variable that can be switched on or off. But when the crowd crosses a historic threshold, there is a portion of variance I cannot explain with any indicator. The right response is to record a confidence interval, not to pretend that portion does not exist.

Another sport follows the same logic: esports. The elite competitive window of a player is shorter than that of a footballer, while youth development and post-retirement support systems are close to non-existent. The public data on those players is thinner even than in swimming. The blank there is larger, and the temptation to fill it is larger too.

Where to start in the next three months

If you ask me where to begin, the answer does not lie in writing more analysis.

The task is to build a single data column for every national final: the 50-metre split. Just one column. Once it exists, every argument about Vietnamese swimming shifts axis, from "who won" to "how did they win, and can that method be reproduced". Nguyen Huy Hoang once put Vietnamese swimming on an Asian Games podium in the distance events. The more interesting question is not how fast he was, but how that speed was distributed across eight laps — and whether another athlete can reproduce that distribution.

A shot happens once. Its trajectory lasts for years.

If next season the national championship organisers publish the splits of every final, who will be the first to open that file?