When a Swimming Analysis Returns a Blank Page

Two in the morning, and a perfectly formatted blank sheet

The report landed at two in the morning, the hour when the worst sports reports tend to arrive. Nine sections. Headings, tables, confidence labels printed in square brackets, even a glossary at the end. A reader skimming it would see a document someone had taken hours to build.

Then I read it closely.

Every content field carried the same line: no information, insufficient data to assess. Not a single athlete's name. No distance, no time, no meet date, no pool. The only populated field was the domain label: swimming. One label, and everything else was white space ruled into neat columns.

Eighteen years of notebooks teach you a reflex that is hard to break: when the page is blank, the writer wants to fill it. I know that feeling better than most. In 2026, at the Asian junior athletics meet in Bangkok, I wrote a long piece built on a single unconventional technical detail: an Indian 400m hurdler named Arjun Singh was running thirteen strides between hurdles instead of fourteen. My editor called it bluffing. Three months later he broke the national record at 48.72 seconds, exactly the odd-stride pattern I had described.

This time was different. There was no odd stride to hold on to. I don't read the record board. I read the lane in their eyes. Here there were no eyes to read.

That is why I sat down to write this. The most dangerous thing in my trade is not a wrong analysis, because a wrong analysis gets caught. It is an analysis typeset so beautifully that nobody bothers to check whether anything is inside it.

A two-stage pipeline and one blockage

Sports desks no longer write with a single pen. A swimming story passes through at least two processing stages before an editor sees it. The first stage deconstructs: it reads the source and returns a list of information points, a one-sentence summary of the author's stance, the entities mentioned, time sensitivity, and a source-quality assessment. The second stage renders those fragments into nine dimensions: technique, performance data, competition system, world landscape, rules and anti-doping, athlete career, risk profile, public narrative, industry ripple.

The report I received was the output of stage two. The problem was in its input.

Every content field from stage one was blank or marked not applicable. And yet one field carried data, and it carried the right one: the domain label, swimming. That inconsistency is the most important clue in the whole episode. A state where the domain classifier ran correctly while the content extractor returned zero is not a state a real swimming article can produce naturally.

A swimming article, even the shortest one, almost never returns zero entities. It has to contain a name. A meet. A distance. A stroke. A pool. Even the driest federation press release holds at least two of those.

So the correct diagnosis is not that the article was empty, but that the extraction step failed while the scaffolding step completed on schedule. It is a pipeline fault, not an article fault.

To analyse swimming you need what swimmers leave behind, and it is a denser data trail than almost any other sport: reaction time off the blocks, the fifteen-metre split, the count of underwater dolphin kicks, turn times at every wall, stroke rate, distance per stroke, and every fifty-metre split. Without those, a writer can only describe feelings.

No lane, no technique

The technical dimension of any swimming analysis stands on five legs: start and underwater phase; turns and finish; efficiency, meaning stroke rate set against distance per stroke; technical progression within a race; and adaptability to venue conditions. The report answered all five with the same sentence: insufficient data.

The answer was honest, and it closed every door behind it. You cannot say an athlete started slowly without a reaction time. You cannot say her underwater phase is weak without a fifteen-metre split.

Based on my experience watching races since my early years as a reporter, the first thing I write in my notebook is never the final time. It is the fifteen-metre split, because it tells you what state of mind a swimmer entered the race with. I believe in intuition, but I have learned to make intuition wait for data. Here intuition had nothing to wait for, and that is a very different condition from being contradicted by data. Being contradicted leaves room for argument. Having no data leaves only silence.

Twenty-five metres and fifty metres are two different sports

Pool condition is a blocking variable. A short-course pool doubles the number of turns, and every turn adds propulsion. The same swimmer, in the same week, in the same condition, can produce times far enough apart to change how a record should be read. Any conversion between the two is reference only, and valid only when both swims have detailed splits to correct with. Without splits, conversion becomes guesswork.

The report stated that pool condition had not been established. It reads like an administrative footnote. In practice it was a freeze order on everything downstream.

An empty coordinate system and the cycle discount

A swimming result only means something inside a three-tier coordinate system: world record, all-time list, and current-season ranking. Without a time, all three collapse together. Behind them sit two questions I always ask: whether the improvement is physiologically plausible, and what the split structure looks like. Neither can be posed without a time, splits and repeat data.

Then comes the four-year cycle discount. A result in a post-Olympic year must be read differently from one in a pre-Olympic year. In post-Olympic years many teams train through meets, accepting poor results to build a base. In pre-Olympic years the same athlete is tapered to peak. From outside, the two swims look identical. Inside, they are two different experiments. The discount can only be calculated with a month and a year, and the report recorded time sensitivity as not assessed.

The most dangerous grey zone is the unclassified one

The field that stopped me longest was anti-doping. The report said simply that it could not determine whether the source article contained any doping element, and it added the most precise sentence in the entire document: unclassified does not mean clean.

The standard professional procedure is to classify immediately into four buckets: a violation confirmed by a competent authority; a contamination dispute; a procedural issue; or a public allegation with no established basis. Four buckets, four legal outcomes, four reputational costs, four ways of writing. Misclassifying in either direction does harm. Filing a proven contamination case under public allegation erases accountability. Filing a public allegation under confirmed violation strips a career from someone who has not been found to have done anything.

Here we do not know which bucket we are in, and that is the worst possible state, because it removes the one thing a writer must do: choose a position that can be defended with evidence.

The only row in the risk table with a number in it

The risk table had six rows. Five said no identifiable risk. The sixth read: systemic risk, level high, probability high, impact high, mitigation — halt publication and re-run the extraction step.

It was the only real judgement in the whole document, and notably it was not about swimming. It was about the process that produced the document. A fully formatted report over an empty payload is a silent-failure hazard: a reader skimming nine tidy dimensions may conclude the source was analysed and found unremarkable. That conclusion is false, and it makes no sound.

If such an output is used to gate a publishing decision, the only defensible action is to hold, not to publish under a low-value label. The two sound similar and differ at one fatal point: a low-value label is a conclusion about the article, while holding is a conclusion about the process.

Contrarian: a beautiful report is more dangerous than a broken one

In eighteen years I have seen many broken reports. A visibly broken one always gets fixed, because it incriminates itself. A beautiful one does not. It looks like what a hurried editor wants to see: evidence that everything has been handled. And because it looks handled, it gets filed.

Here lies the heart of it. No signal is different from low signal. Low signal is a finding: we looked and found little. No signal is an incident: we never looked. Systems routinely merge the two because both produce a number near zero. Professionally they demand opposite actions.

The second temptation is worse: filling the gap with general knowledge about swimming. I could write three thousand words on turn mechanics without a single fact about a specific swimmer. It would read smoothly. It would be wrong in no sentence. And it would be a complete fabrication, because it would present itself as case analysis while being a lecture in disguise.

This is the characteristic failure of an era in which text became cheap. When prose costs nothing, the scarce good is verification. A fluent article over an empty factual base is worse than a clumsy one over a full base, because the first takes from the reader something they cannot recover themselves: the belief that someone checked.

I once went against that principle and was lucky. In March 2026, when every meet was cancelled and stadiums stood empty, Arjun Singh tore a hamstring and lost his Olympic place. I wrote something bleak. Two weeks later I watched a short clip of a twenty-two-year-old Kenyan named Wanjiru, training alone on a dirt road in Thika, no coach, just a stopwatch and a notebook. I flew there on my own money and stayed eight days.

What made me go was not a feeling. It was two objects visible in the frame: the stopwatch and the notebook. An athlete timing herself and writing it down is an athlete building her own programme. That is an observable, checkable detail. A crisis takes away the stadium, but it does not take away the trajectory. In that report, the crisis had taken the lane itself.

There is one more contrarian layer few in the trade say aloud: the person who holds is rarely rewarded. A desk that measures output by article count gets nothing to count from a decision to stop. It only creates a gap, and the gap gets filled by someone willing to write on instinct. That incentive structure explains why empty reports keep going out the door, and why they are so rarely stopped.

The finish line is one verifiable name

The hurdler does not ask how high the hurdle is. They ask where the route to the finish lies. For that report, the route lay in something very modest: an athlete's name, a distance, a competition date, a pool length. Four fragments, enough to reopen every door.

What is worth keeping from this is not a lesson about swimming. It is a lesson about gates. Any analytical pipeline needs a minimum condition before it is allowed to produce conclusions: the list of information points must contain at least one element. A gate so obvious it is usually absent.

As sport generates data faster than anyone can read it, the marginal value of journalism is not another paragraph. It is knowing where not to write. A well-timed stop is an editorial act, not an evasion, and newsrooms should treat it as a result rather than a hole to be filled.

Next time a pipeline returns a blank page ruled into neat columns, the question will not be what we write from it. The question will be who picks up the phone to re-run the first step, and whether the desk has the patience to wait for a real name.