The Empty Cell on the Results Sheet: When Athletics Data Is Missing, Conclusions Must Stop

A night of competition at the National Stadium in Tokyo, and the results sheet printed in the press room has twelve columns. Rank, name, nationality, time, reaction time, personal best. The wind column is blank. Fifteen minutes later a video appears online with the words “new record”, and in our analysis room group chat, seventeen people type almost simultaneously: what was the wind?

Nobody can answer. The sheet is not wrong. It is structurally complete, correctly formatted, with enough fields to feed a model. Only one cell has no value. And that cell decides whether the rest of the page means anything at all.

I have worked in sports data analysis in Tokyo for five years, since I left the track. My job is to read sheets like that one before they reach the newsroom. Watching competition live has taught me an uncomfortable conclusion: most mistakes in athletics analysis do not come from reading the numbers wrongly. They come from reading an empty cell as if it were a zero.

An empty cell and a zero are two different quantities, and the gap between them is where an entire analysis industry fools itself.

A complete sheet is not a correct sheet

In any data system, a record can pass every structural check and still be worthless. It has all the fields, all the right data types, all the right date formats. It simply has no content. Engineers call this a silent failure: the system raises no error, the user sees no warning, and the final product still ships looking perfect.

Athletics produces silent failures more easily than most sports, because it rests on a beautiful belief: the number on the board is objective. The stopwatch has no bias. The tape measure has no side. But that objectivity applies only to what was measured. What was not measured is neither objective nor subjective — it simply does not yet exist in the frame of reference.

A major championship season is when silent failures do the most damage. Newsroom pressure multiplies, the audience reading results multiplies by a hundred, and the time available for cross-checking shrinks to minutes. One empty cell that slips through that window will be replicated into thousands of articles, millions of views, and a generation of false memory about a race.

For five years my working process has revolved around a single question: what is missing from this sheet? Before asking what the numbers say, I ask whether the sheet is complete enough to say anything. It looks slow. It is far cheaper than repairing a wrong conclusion later.

The wind column: the +2.0 m/s threshold and what it erases

World Athletics sets the maximum assisting wind for record ratification at +2.0 metres per second in the sprints, long jump and triple jump. Above that threshold the mark remains a valid competition result, still ranked, but cannot be ratified as a record. It is a small technical rule, and it is the only barrier between “a very fast athlete” and “the fastest athlete in history”.

When the wind column is blank, both sides of that barrier collapse. Nobody can confirm, nobody can deny. In practice, though, the default public response is confirmation — because a record is a better story than an ordinary mark, and the better story is always told first.

I have seen this at a domestic meet. A school athlete jumped far beyond her previous best, the board showed the mark, the stand erupted, and three newspapers ran the story that same evening. The next morning the wind gauge was checked and found not to have recorded at all during that event. No record was stripped. There was simply a mark that was never confirmed. But the celebration had already been broadcast.

In Vietnam's jumping events this is a familiar blind spot. Bui Thi Thu Thao once put Vietnamese women's long jump on the regional map, and every time a strong mark appears, my first question is not how many metres but how much wind was recorded. At meets with inconsistent measurement equipment, one empty cell can turn a regional result into a continental claim inside a single headline.

Altitude: the Mexico City lesson is still only half-read

In October 2026, in Mexico City, Bob Beamon jumped 8.90 metres. The city sits more than 2,200 metres above sea level. Thinner air, less drag, and a large number of world records fell during the same Olympic Games. Beamon's mark survives to this day, but the way it survives always carries a note about altitude.

That note does not exist to diminish the athlete. It is the condition that makes comparison possible. A mark at 2,200 metres and an equivalent mark at sea level are two different physical events, and the world rankings do not reflect that difference.

In my work I always place altitude before any other reading. The same athlete, the same month, the same training load, running in two different locations can produce two results far enough apart to change the entire assessment of form. Ignore that variable and my model is not wrong because the algorithm is weak. It is wrong because I fed it an incomplete input.

The 40mm shoe and the dividend nobody deducts

In 2026 World Athletics introduced a rule capping road shoe sole thickness at 40mm, together with a requirement that shoes be available on the market before competition use. The rule arrived after a run of distance marks drew suspicion about carbon-plated technology. It was a rare moment of the sport publicly admitting that part of a performance comes from equipment, not only from legs.

Admitting it at the regulatory level is not the same as deducting it in analysis. Almost none of the comparison tables I receive carry a shoe column. People compare a 2026 marathon time with a 2026 marathon time as if both athletes ran on the same surface and the same technology. That technology dividend is left behind the numbers, and it does not disappear on its own.

Tigist Assefa ran 2:11:53 in Berlin in 2026. Eliud Kipchoge ran 1:59:40 in Vienna in 2026 in a purpose-built event, and precisely because of how that event was staged the mark was never ratified as an official world record. The two numbers sit side by side in a great many articles. They sit side by side in no frame of reference whatsoever.

Reaction time and the splits that get dropped

A 100m result contains at least three layers: total time, reaction time, and 10-metre segments. News coverage keeps the first layer, because it is the only one that fits a headline.

When Noah Lyles won in Paris in 2026 in 9.79, most spectators learned exactly one number. To assess real form, an analyst needs to know where he accelerated, at which metre top speed occurred, and whether his reaction sat inside his own normal range. Marcell Jacobs won in Tokyo in 2026 in 9.80 in a season where he had almost no stable performance base beforehand — a case where reading only the final time produces a completely wrong conclusion about his standing in the sprint world.

Drop the splits and a model treats every athlete as a single pixel. It loses the ability to separate the athlete who runs fast evenly from the one who is slow but unleashes over the final 60 metres. Those two athletes have entirely different career longevity, injury risk and racing strategy. Dropping splits means dropping the ability to see the difference.

Qualification architecture: two pathways and three slots

At Olympic and World Championship level, entry runs along two parallel paths: hitting the entry standard, or accumulating world ranking points inside the qualification window. Alongside that sits a cap of three athletes per country in most individual events.

Those figures create a problem the results sheet cannot show. An athlete ranked fourth domestically may have a better mark than another country's champion, and still stay home. Conversely, a one-race-decides-everything selection system such as the US trials creates the opposite risk: the fourth-place finisher at a national trials can be a genuine world medal contender who never boards the plane.

For Vietnamese athletics the story is more layered still. Nguyen Thi Oanh won three individual gold medals at the 32nd SEA Games in 2026, two of them on the same day. Seen as a performance sequence, that is astonishing. Seen through qualification architecture, it is a resource-allocation problem: one athlete, three events, one finite recovery budget, and a schedule not designed for a multi-event runner.

Nguyen Thi Huyen in the 400m hurdles is another example of the same principle. In that event a hundredth of a second can correspond to several places in international ranking, which makes every non-performance factor — schedule, number of qualifying rounds, recovery between rounds — a genuine tactical variable.

The age curve and the trap of a single season

Athletics has a feature football does not: a fairly clear age curve that differs by event group. Sprints peak between roughly 24 and 29. Distance events peak later, around 26 to 31. Throws peak latest, around 28 to 33. These are broad reference bands, not laws, but they are enough to surface anomalies.

An explosive season at 31 in a sprint event is a signal requiring cross-checking, not a signal requiring immediate celebration. A performance jump far beyond years of momentum is a signal to be placed next to injury history, coaching changes and equipment changes. It does not mean there is a problem. It means there is a question.

When age data and personal-best series data are both missing, there is no way left to separate a rising athlete from a declining one. The default media handling is to treat everyone as rising, because that is the easier story to tell.

Anti-doping: “no finding” is not “no risk”

Modern anti-doping runs on the biological passport, whereabouts obligations, and long-term sample storage — ten years under World Anti-Doping Agency rules. All three share one trait: they operate backwards through time. A result today can be reviewed next year, and a medal can change hands years after the ceremony ended.

So when an analysis raises no doping issue at all, the reader must separate two situations. First: it was checked and nothing was found. Second: there was not enough data to check. The two look identical on paper and differ completely in meaning.

This is why I always state the verification status in every report. “Not assessed” is an honest label. “No issue” is a conclusion that requires evidence. Confusing those two labels is the fastest route to turning an analysis into a warranty it has no authority to issue.

When the data speaks

In 2026, when German football returned to empty stadiums, I collected the first 26 matches and found home advantage falling from an average of 0.44 goals per game to 0.15. Same league, same clubs, same players. Only the context had changed. The empty summer taught me that an empty seat is also a player. The results sheet is the same: an empty cell is also an indicator.

That lesson transfers intact to athletics. A number says nothing on its own. It speaks only when it stands beside wind speed, altitude, shoe type, competition type, and its position within that athlete's own performance series. Remove one piece and the number does not weaken — it becomes something else, and that something else is usually read as a conclusion.

When the data speaks, the laughter is only noise. But that sentence holds only when the data actually speaks. An empty cell does not speak. It stays silent, and silence is always filled by the loudest person in the room.

The contrarian angle: a complete dataset is more dangerous than a broken one

The natural reflex is to trust the sheet with more columns. The more detail, the safer it feels. In operational reality, however, a dataset that is entirely broken is safer than one missing exactly one cell. Everyone sees a broken dataset and stops. A dataset missing one cell still runs, still prints, still gets read, and nobody knows they are reading a fragment.

In the meeting room, emotion asks and data answers. But data can only answer questions it has inputs for. When inputs are missing, the only correct answer is to postpone. Sports analytics rarely chooses to postpone. We interpolate, we estimate, we say “roughly, from experience”. Every time we do that, we create a second data layer with no provenance, and that second layer gets cited as though it were the first.

There is one more thing I learned after many mistakes: emotion is not noise. When a record is doubted, when an athlete is mocked, when a stand falls oddly silent — that is data, merely unlabelled data. Every jeer is an unlabelled data column. The analyst's job is to label the column, not delete it.

And this is the hardest part. I do not guess athletics; I measure the distance between expectation and the track. But to measure a distance I need both ends. When one end is missing, the measurement does not return zero. It returns a gap, and a gap is always filled with whatever people want to believe.

What to watch in the next cycle

The signal I will track next season is not the number of records broken but the null rate on published results sheets. That is a data-quality indicator, and it forecasts analytical quality better than any model.

Every time I open a results sheet, I will ask the old question before the new one: is this sheet complete enough to speak? If not, the most honest answer is to leave it open. Humility before randomness is not an evasion of conclusions. It is opening the door to the possibility that a variable is missing — and that variable may be sitting in the empty cell on the page.