The Empty Cell in Swimming Injury Records: The Cost of a Conclusion Without a Source

In a shoulder-injury monitoring file for a group of swimmers aged 13 to 15 that I was asked to analyse earlier this year, one cell held me longer than all the others. It was the column for accumulated training volume over the 14 days running back from the date the athlete first reported shoulder pain. The cell was blank. Nobody had ever written in it — not written wrong, not written short. Simply left empty. In that same file, 11 of 17 shoulder-pain cases had a blank volume cell. The most complete column was symptom, with short lines such as “right shoulder pain, hurts overhead,” “rested 3 days,” “swam again.” The second most complete column was conclusion. The report sent to the coaching staff read: “Cause: breaststroke technique not yet correct, right shoulder under excessive rotational load.” Four lines of raw data, one causal verdict. At Lach Tray I learned to read injuries from the first numbers. The first lesson, and the least comfortable one, is this: when the first number does not exist, everything written after it is guesswork dressed in clinical vocabulary. I work as an injury analyst specialising in swimming. My job is to build training-load monitoring systems for swim squads and then read those systems to answer two questions: where is this athlete on the load curve, and how long until the body answers back. A decent swimming injury file needs two layers. The capture layer records what happened in each session: metres, session count, stroke distribution, equipment used, perceived intensity, pain site, onset date, history. The analysis layer takes what layer one produced, builds a hypothesis, benchmarks it against the group baseline, and then eliminates. Layer two does not generate data. It only rearranges it. Last month I received a document that was purely layer two: a nine-dimension deep analysis framework for the swimming domain, complete with technical tables, performance tables, a risk matrix, competition-cycle analysis, a doping-governance section and an industry ripple map. A beautiful frame. But when I opened the input section, layer one returned nothing: no article title, no athlete, no event data, no figures at all. What matters is that the framework answered correctly. At every data-dependent position it wrote “insufficient information to assess” rather than inventing an athlete, a result or a story. That is correct behaviour for an analytical system — and it is behaviour Vietnamese sports medicine rarely permits itself, because we are pushed to produce conclusions before the data is ripe. An empty cell has a strange pull. Nobody leaves it alone. In injury records a blank is usually filled with four familiar materials: curse, mentality, luck, and innate constitution. All four share one property — they cannot be verified and cannot be refuted. They are perfect for filling space. In swimming, the most common filler is technique. Bad technique is the default answer for almost every case of shoulder, back or knee pain. It sounds plausible, and it is almost always partly true, because everyone’s stroke has something imperfect. A conclusion that is always partly true stops being a conclusion; it becomes a sentence that cannot be wrong. Subtraction is still the best tool I have. Kane in 2026 was not a curse, it was simple subtraction: remove luck, remove mentality, remove timing, and what remains is an overload problem. In swimming that subtraction needs a minimum of five data fields, and if any one is missing the subtraction does not run. The first field is accumulated volume. Not the weekly average, but total metres over the last 14 days set against the baseline of the previous four to six weeks. In swimming, total metres is the crudest and strongest variable, because it multiplies directly with the number of times the shoulder rotates overhead. A swimmer doing 25,000 metres this week against a six-week baseline of 18,000 is in a different load zone from a swimmer doing 25,000 metres all year. Same figure, two different risks. Without the baseline the figure means nothing. The second field is stroke distribution. Shoulders load differently in butterfly, freestyle, backstroke and breaststroke; knees load differently in breaststroke and freestyle. A week in which butterfly share spikes while total metres stay flat is still a load spike — it just does not appear in the total column. The third field is equipment. Fins, kickboards and especially hand paddles change drag force and shoulder angle at entry. In many junior programmes paddles are used as a reward for technical progress, and nobody records the day they were introduced. It is a confounder that gets forgotten constantly. The fourth field is pain history. Across most injury-surveillance literature, previous injury is the strongest predictor of the next one — stronger than age, stronger than volume. A swimmer who had shoulder pain last season enters this season with an already elevated risk floor, and any analysis missing that field is discarding its strongest variable. The fifth field, for girls aged 12 to 16, is biological stage. Height, arm span, fat mass and centre of gravity shift fast enough in this window that technique learned last season becomes misaligned this season. Leaving this field out of the file removes exactly the variable that explains most of the shoulder pain appearing suddenly at this age. Those five, plus the competition calendar, form the minimum threshold for a grounded injury conclusion. Below it, every answer is prose. The field already has a handsome tool for load: the acute-to-chronic workload ratio. It became popular around 2026 and was used for a while as a standard indicator. Then methodologists turned on it, arguing that a ratio between two already-correlated quantities can manufacture spurious correlation, and that a tidy index does not equal a causal relationship. I followed that argument for years, and what I took from it is not whether the index is right or wrong. It is that the tool is not the data. A beautiful index built on empty cells is still an empty cell with a border drawn around it. Swimming has a tradition worth borrowing here. In official competition results, blanks exist legally. A swimmer who cannot start is recorded as DNS, Did Not Start. A swimmer disqualified is DSQ. A touchpad that fails to register is NT, No Time. Nobody in the swimming world fills in a number to make the sheet look nice. An empty cell in a results sheet is part of the result. And yet in injury records the same culture accepts filled numbers. A shoulder case with no onset date, no baseline volume and no history still gets a causal verdict written next to it. The difference is this: results sheets have officials, and injury records do not. Three ways blanks get filled, most often. The first is filling with performance: a swimmer competing below form is explained by a fitness slump, while nobody holds the training metres from the three weeks before the meet. The second is filling with character: she is lazy, he is mentally weak. The third is filling with collective story, and this is where curse narratives are born — the cursed lane, the cursed meet. Empty stands, a golden rule bent, and the body pays. The period when pools closed and reopened with compressed calendars is the clearest case I have recorded. When the ten-day progressive load protocol was cut to three days to make a meet, the rise in shoulder and back pain did not sit with the squad training the most. It sat with the squad that returned earliest. Some gaps can be patched. Without metres, session count multiplied by perceived intensity can build a proxy — provided you label it a proxy and never compare it with another group’s baseline. Without stroke distribution, you can at least record how many sessions used paddles. The gap that cannot be patched is the onset date. Lose the onset date and there is no way to tie symptoms to a specific load, and every causal conclusion from that point is reverse inference. Every fall has a graph, and every graph has a break point. Swimming has no falls. It has load curves, and every load curve has a break point — it only appears when somebody bothers to draw the line. Here I have to argue against myself, because there is a mirror version of this position that easily becomes a shield. If insufficient information becomes a permanent answer, the analyst never has to be accountable for any conclusion. That is another kind of fabrication: fabrication by silence. In Vietnamese swimming, most small centres keep records in notebooks, and if I demand a complete data system before clearing a swimmer to return, the practical outcome is that nobody records anything. Format perfection kills capture. Rough, handwritten, incomplete data still beats a blank sheet designed to standard. What I propose is a deadline. Three minimum fields — baseline volume, onset date, prior history — must exist within 72 hours of an athlete reporting pain. After that mark, a conclusion is allowed, with the uncertainty level stated in plain words. That is different from waiting for beautiful data, and different again from concluding on day one. The execution blind spot is elsewhere, and it is not with the coach. It is with the analyst. We tend to build the framework first and then go looking for data to fill it. A grand nine-dimension frame creates the feeling of serious work, when the thing that needed doing first was a five-column page, filled completely for three straight weeks. If a 13-year-old swimmer of yours reports shoulder pain today, how many numbers about her last 14 days do you actually hold? If the answer is fewer than three, you are not yet looking at an injury that needs diagnosing. You are looking at an empty cell waiting to be filled. The body is a closed system, but data is the key that opens it. And that key can only be forged one way: by recording what actually happened, even when nobody has asked for it.