Empty Cells on a Swim Results Sheet: The Discipline of Not Guessing

At a time-trial session in a 50-metre pool in Nha Trang, a coach placed a results sheet in front of me with twelve empty cells. He asked one short question: “So did he swim well or not?” I told him the sheet was not enough to say anything at all. He looked at me as though I had just refused to work, then carried the sheet over to someone else who was willing to deliver a verdict in thirty seconds.

I remember that afternoon clearly, because it is a miniature of the entire sports-data profession. Every empty cell is an invitation to fill it in. The first empty cell is reaction time. The next is the 15-metre split after the start. Four cells after that are 25-metre splits. Two more are turn times at the wall. The rest are stroke rate and distance per stroke. Without them, a results sheet becomes a number without coordinates.

In swimming this happens more often than outsiders assume. Swimming is a sport where every metric is generated by the athlete alone, with no opponent blocking, no ball, no dispute-induced error. It sounds clean. But precisely for that reason, gaps are easier to fill with narrative than in any other sport. A sheet that contains only a finishing time will always leave room for a compelling story.

Swimming does not lack data. It lacks data in exactly the places that decide whether a conclusion is right or wrong. Finishing time is the easiest thing to measure and the least informative. It is the sum of everything that happened before it, and because it is a sum, it conceals everything that happened before it.

I learned this lesson a different way, in a different sport. In 2026, when pitches closed because of the pandemic, I sat down with five V-League seasons and two hundred and forty players in a spreadsheet. I had no new video, no new matches, only old data dense enough to reveal patterns. COVID shut the stadiums, so I reopened the V-League directory. No league is meaningless. The result of that process was a finding about Nguyen Trong Hung: he was still scoring, but his acceleration had dropped 38 per cent from the previous season. I wrote that he would fade after the 70th minute and advised the club against extending his contract. The club leadership replied that I should not sit in Nha Trang talking about what happens on a pitch. When football returned, he moved to Binh Duong, played 11 matches and lost his starting place.

What I carried from that story into the pool is not the conclusion itself, but the mechanism. Goals are an aggregate metric. Acceleration is a structural metric. When the aggregate holds steady while the structure decays, the media keeps praising and the model has already recorded the decline. In swimming, the equivalent pair is finishing time and the first 15 metres. An athlete who holds his personal best while his starting speed steadily falls is compensating with endurance, and endurance is an asset with an expiry date. The final time is only a sum; the opening speed is the fingerprint.

Back to the sheet with twelve empty cells. Suppose I were forced to conclude. I would write: the athlete finished in 2 minutes 04.8 seconds. It sounds tidy. But if the first 100 metres was 59.6 and the second was 65.2, that is a fade, not a sprint finish. If the two figures are 62.4 and 62.4, that is an even swimmer. If they are 64.0 and 60.8, that is someone who paced himself. Three different structures, one identical result, three different reports. The audience receives a single line, and that line cannot tell who is who.

Numbers never lie, but they know how to hide.

There is a second, subtler form of gap: data that exists but does not share a frame of reference. A 25-metre pool and a 50-metre pool produce the same figure with entirely different meaning, because the number of turns differs. Automatic timing and manual timing produce different margins of error. Heats and finals, individual races and relays, are contexts that cannot be compared directly. A ranking that mixes all of these creates the illusion of a verified order.

In 2026, I spent three days rewatching Germany’s entire World Cup group stage. Against South Korea they controlled 74 per cent of possession, but their PPDA rose to 13.2, meaning they allowed the opponent 13 passes before engaging. Their forwards ran an average of 6.3 kilometres per match. Germany 2026 did not collapse through luck. PPDA had announced it from the group stage. The aggregate said one thing, the structure said another, and the structure was right.

The same logic applies to the pool. In 2026, I recalculated Patrik Schick’s entire European campaign: five goals, one of them struck from 49.7 metres, while his total expected goals for the tournament was just 2.6. That long-range strike carried an xG of 0.03. He sat outside any sustainable prediction band, and his transfer value was inflated by moments that do not repeat. In swimming, the equivalent is a personal best standing alone, far from that same athlete’s median across his last twelve swims. One swim is news. Twelve swims are ability. Luck is something I do not have. I have probability and data thick enough to work with.

This is where I have to argue against myself. People assume that a data analyst who refuses to conclude is weak at the job, lacks nerve, or is hiding something. The opposite is true. Refusing to conclude when the data is insufficient is a deliverable, not an evasion. An empty cell left empty forces an organisation to go back and measure. An empty cell filled with a guess becomes a salary, a contract, and a competition slot belonging to a human being.

But I have no interest in building the image of a man who has never been wrong. Based on my experience following competition across many seasons, I once declared that a young athlete would break a national record after watching exactly two swims, and I was wrong. Two swims are not a sample. I publish that mistake as often as I publish my correct calls, because posting only the hits means selling belief rather than analysis. Correlation is not causation, and a beautiful trend line does not automatically become a valid forecast.

So when a sheet with twelve empty cells lands on my desk, the first thing I inspect is not the athlete. I inspect the dataset. Which column is empty, and is it empty because the measurement system does not capture it, or because the person entering data skipped it? If it is missing because of the system, that is an investment problem. If it is missing because of the operator, that is a discipline problem. Two entirely different problems, with different costs and different fixes.

Over the next three months, the signal I am waiting for is not a new record. I am waiting to see which domestic meet publishes reaction time and 15-metre splits as a default part of the results sheet, instead of a single finishing number. The meet that does it first takes ownership of how its own performances are explained, rather than leaving the story to someone else. Teams do not collapse in one night. They collapse when their metrics stop connecting to each other. The same holds on the pool deck, except there nobody covers for you.