Athletics and the Data-Integrity Problem: Why an Empty Sheet Can Be More Honest Than a Fabricated One

ATHLETICS AND THE DATA-INTEGRITY PROBLEM: WHY AN EMPTY SHEET CAN BE MORE HONEST THAN A FABRICATED ONE

An empty file at 11:47 p.m.

At 11:47 p.m., in a small apartment in Nishinari, Osaka, I opened a data file I had been waiting three weeks for. It weighed twelve kilobytes. Inside were nine sections, and every one of them ended with the same phrase: insufficient information. No competition name. No athlete name. No mark, no wind reading, no timestamp. A fully built analytical frame with nothing inside it.

Outside the window, a Nankai train crossed the bridge and its yellow light swept across the ceiling. I remembered Rostov-on-Don in July 2026. On that night in Russia, I watched the data shatter in front of me. I was seventeen, sitting in front of a screen with a handwritten notebook, counting every touch of the ball by the Japanese national team against Belgium. This time there was no one to blame but the process itself.

The silence of an empty file weighs more than the silence of a lost match. A lost match still leaves data behind. An empty file leaves nothing but a question: what happened at the input stage.

From Rostov-on-Don to Yodoko Sakura

My job is to read numbers, and I arrived at this job through a mistake that the internet attacked me for. That night in Russia, Japan led Belgium by two goals and lost 3-2, with the final goal arriving in the fourth minute of stoppage time. I recorded one figure that still holds value eighteen months later: Japan made only seven touches inside the opponent's penalty area, while Belgium made twenty-one, even though Japan held roughly 55 percent of possession. I wrote a post on my personal blog concluding that pushing the defensive line high in the final minutes was a structural error.

A group of supporters reacted fiercely. They said I did not understand football, that the emotion of a match could not be reduced to a spreadsheet. I held my position, not because I was certain I was right, but because I knew how to re-check my own numbers.

In 2026 I joined Runner's World as editor-in-chief, writing thousands of pieces about running. That was the period when I learned a simple discipline: every article must begin with the question of what the data says, not with the question of what I feel.

In the spring of 2026, the J-League postponed four months of fixtures. I was a journalism student in Osaka and could not go to Yodoko Sakura Stadium to watch Cerezo Osaka. I stayed home and rebuilt a dataset from old match footage, logging 1,240 pressing situations from Cerezo's 2026 season in order to calculate the number of passes the team allowed opponents before closing them down.

From that dataset I predicted Cerezo would decline once the league resumed, because losing home matches meant losing part of the pressure placed on opponents. The result: Cerezo finished fourth, below the second place my model predicted. I published a note admitting the error and added a new variable to the model, which I called crowd effect. An empty stadium, but the numbers were still full of noise.

In the summer of 2026, I spent three weeks tracking the European Championship and found that Denmark scored four of six goals from pre-designed set pieces, far above the tournament average. I placed that data beside the set-piece record of RB Leipzig in the 2026-21 Bundesliga season, where coach Julian Nagelsmann used running-position and ball-landing data to design training drills. Every corner kick is now a mathematical proposition. My article was republished by a small football site and reached 15,000 reads.

In January 2026 I worked as a contributor for an online football magazine during the winter transfer window. I analysed data on more than two hundred players moving from the J-League to Europe and found a correlation coefficient of 0.67 between kilometres run per match in Japan and success rate in the Bundesliga. I contacted a scout at a German club and recommended midfielder Ao Tanaka, who was running 11.8 kilometres per match, the highest in the J-League at the time. He moved to Fortuna Düsseldorf on loan. The contract is only the ending; the beginning is in the spreadsheet.

But every story above rests on a single assumption: that the data exists. When I moved into athletics coverage for the Japanese market, I realised that assumption collapses far more often than it does in football.

Athletics is a sport of hidden error margins

Football hides its data behind hundreds of actions per match. Athletics does the opposite: it exposes a single number, and precisely for that reason, every impurity around that number becomes more dangerous. One athlete runs 9.79 seconds and another athlete also runs 9.79 seconds, and they can be separated by five thousandths of a second. In the men's 100 metres final at Paris 2026, Noah Lyles and Kishane Thompson both finished with 9.79, and the gold medal was decided by photo finish. That margin is smaller than the timing error of the equipment used on many national-level tracks.

That is why the input check matters more than the conclusion.

The first variable: wind

World Athletics recognises a record only when the tailwind does not exceed 2.0 metres per second at the moment of competition. A 9.79 run with a plus 2.0 wind and a 9.79 run with a minus 0.3 wind belong to different tiers, even though the scoreboard displays them identically. In many analyses I read online, the wind reading is removed entirely. The writer compares two numbers as if they were equal and draws a conclusion about form. That is not analysis. That is copying.

In Asia, I frequently see summary tables of Southeast Asian athletes' performances that mix valid wind readings with readings above the threshold. Reclassifying takes about twenty minutes for a single meet. Nobody does it, because the reclassified result is usually less attractive than the original.

The second variable: altitude

Mexico City sits above 2,200 metres. Bogotá exceeds 2,600 metres. In those places, thinner air reduces drag, and sprint or long jump events produce marks that cannot be reproduced at sea level. An athlete whose personal best was set in Bogotá is not cheating, but that athlete's development curve will look distorted on paper if the analyst does not note the location.

I once saw a Japanese sports outlet call a young long jumper a phenomenon after a meet in South America, without mentioning altitude. Three months later, the same athlete competed in Japan and jumped nearly seventy centimetres less. Nobody explained why, and the audience assumed the athlete's form had dropped.

The third variable: shoes

From 30 April 2026, World Athletics imposed a sole thickness limit of 40 millimetres for road events and 25 millimetres for track events, and required that a shoe be available at retail for at least four months before competition use. The rule arrived after a period of record surges linked to a generation of shoes with embedded rigid plates.

What matters is not the 40-millimetre figure. What matters is that historical ranking tables still mix results from two different technological eras without a note column. A national record set in 2026 and a national record set in 2026 appear on the same line, as if two athletes had run under identical physical conditions.

The track at the Tokyo 2026 Olympic Games was designed to optimise rebound, and the summer of 2026 produced a wave of world records in hurdling events. When analysing an athlete's marks from that period, I always separate two data groups: results on the new track and results on a standard track. Without that split, every comparison is meaningless.

The personal best curve and the trap of youth

A personal best is not a point; it is a curve. When I receive an athlete's file, the first thing I do is redraw it year by year. A steadily rising curve tells a different story from a curve that spikes once and then flattens.

Sprint events usually peak between ages 24 and 27. Marathon events usually peak later, between 28 and 33. If a 21-year-old has already reached their highest mark, that is a signal to monitor, not a signal for the media antenna to celebrate. Conversely, a 31-year-old still improving a personal best over 10,000 metres is a rare case, usually linked to a change in training method rather than luck.

Alongside the performance curve runs the training-load curve. Most athletics injuries do not come from a single collision; they come from accumulated load exceeding the tolerance of tendons and bone over a long period. Training-load data is almost never published. That is the largest blind spot in this sport.

When an athlete unexpectedly withdraws from a major meet, the media usually reports the injury as a sudden event. To me, it is the output of a process whose data existed somewhere, and nobody chose to publish it.

The Olympic entry gate is not a single door

The qualification mechanism is one of the most misunderstood parts of the sport. A qualifying standard is only one of several paths. For the Paris 2026 Olympic Games, the marathon entry standard was 2 hours 08 minutes 10 seconds for men and 2 hours 26 minutes 50 seconds for women, applicable within a defined window. Athletes who miss the standard can still earn a place through the World Athletics ranking, where points accumulate from finishing positions and performances at meets inside the system.

Beyond those two routes there are universality places for countries without representation, and decisions made by national federations. Japan runs a dedicated selection race for the marathon, where the race result itself decides the place, before any additional slots are allocated. That system means a national-record holder can still miss the Olympic Games by failing on one specific day.

For an analyst, this is the easiest place to make a mistake. We tend to assess an athlete's chance of qualifying from personal bests, while in reality the chance depends on the competition calendar, the number of starts inside the qualifying window, and the federation's meet-selection strategy.

A power map with cracks

The world athletics map has been stable for decades. Kenya and Ethiopia dominate distance events. Jamaica and the United States dominate sprinting. Japan stands out in the marathon and in long-distance relays, partly thanks to an ekiden culture of large-scale relay races in early January each year, which produces a stream of young athletes conditioned for endurance from an early age.

But cracks are appearing. Botswana won the men's 200 metres at Paris 2026 with Letsile Tebogo, opening a new sprint axis in southern Africa. European middle-distance running revived through Jakob Ingebrigtsen and Josh Kerr, and at Paris 2026 it was Cole Hocker who won the men's 1,500 metres, while Ingebrigtsen finished off the podium.

To me, the most interesting detail in that picture is not who won, but the depth behind the winner. A country with one champion is a country with a star. A country with five athletes in the same final is a country with a system. I assess a nation's athletics strength by the number of athletes meeting the standard in each event, not by medal count.

The grey zone of the rulebook

No sport has a legal system as complex as athletics, and that makes rule analysis a mandatory part of the job.

Under World Anti-Doping Agency rules, three whereabouts failures within twelve months counts as an anti-doping rule violation, even where an athlete has no positive test. The athlete biological passport, in operation since 2026, allows abnormal blood variations to be detected over time instead of relying on a single sample.

In another direction, regulations covering female athletes with differences in sex development create a grey zone that is both scientific and human rights related. The case of Caster Semenya has passed through multiple international jurisdictions, from the Court of Arbitration for Sport to European human rights institutions. Analysing results in affected events without stating the applicable legal framework is a professional error, because the legal context determines how the number should be read.

I have no ambition to conclude who is right or wrong in those disputes. The analyst's job is to state clearly which legal framework applied when a mark was set, so the reader knows what is being compared.

Training systems and peaking cycles

In Japan, sports science support is concentrated in a number of national research centres and sports universities, where athletes have training load measured, sleep tracked and nutrition adjusted by cycle. This model contrasts with the oral-tradition approach that remains common elsewhere.

Peaking is a concept that gets abused. Reaching top form on one specific day of the year requires load distribution accurate to the week, and a small deviation in the taper phase can ruin an entire season. When analysing an athlete preparing for a major meet, I do not look at their best mark of the season, but at the gap between their best mark and their average mark over the final six weeks before the meet.

The smaller that gap, the higher the probability of a stable performance at the meet. It is one of the few predictive indicators I can calculate from public data.

Risk matrix and priority order

In the risk table I apply to every analysis, the top item is not doping, not injury, but data integrity. The reason is simple: every other risk is assessed on the basis of data, and if the data is wrong at the root, the entire assessment downstream is just decorative scaffolding.

My priority order has six layers. Layer one is data source and traceability. Layer two is completeness of measurement. Layer three is adjustment variables, including wind, altitude and equipment. Layer four is the legal context. Layer five is doping and injury risk. Layer six is media risk.

When layer one fails, I stop. There are no exceptions. An empty analytical sheet is more honest than a sheet full of unverifiable numbers, because a full sheet creates the illusion of knowledge.

Public narrative and the expectation trap

Public attention runs in cycles. A young athlete who breaks a national record will be called a phenomenon. Two months later, if that athlete does not improve, the media starts asking about stagnation. Most of those cycles are based on no actual change in ability, but on the content demand of the media system.

I measure this with three indicator types: the volume of content produced about an athlete over a period, the ratio between performance language and emotional language in headlines, and the gap between public opinion and underlying data.

The gap between market expectation and objective assessment is a useful indicator. When that gap is large, the probability of a result shock is also large. Every probability contains a shock; I only make sure it does not repeat inside my model.

Transmission into the industry

The reach of an athletics result does not stop at the track. Competition shoe technology is the clearest example: a rule about sole thickness can change an entire sports brand's product catalogue within twenty months. Next comes the personal sponsorship market, where an athlete's contract value depends on sustaining performance across multiple seasons.

Further out is the youth development system. When a country has athletes meeting Olympic standards, the number of children registering at athletics clubs in that country usually rises over the following one to two years. This effect is measurable, and it matters more than a medal, because it shapes the athlete supply of a decade later.

Why I leave the cells empty

I have had the chance to fill an analytical sheet with speculation. There were nights when I knew that adding one name, one meet, one plausible number would make the article read far more smoothly. I did not do it, because the lesson from Cerezo Osaka in 2026 is still intact: when I added the crowd-effect variable to the model, the error shrank but the model became more complex, and I had to rewrite the entire conclusion section.

I collect mistakes, classify them, and then I know where a team is heading. That method is slow, and in this profession slow is rarely rewarded. But data does not create stories; it strips the stories of others bare.

Someone three years into the job can write a very fluent piece about an athlete without verifying anything. Someone nine years into the job knows that what gets lost is not one article, but the reader's trust when the first number is found to be wrong.

In Japan, I work in an environment where accuracy comes before speed. In Vietnam, where I was born, supporters receive information differently: they forgive errors if the writer is honest, and they turn away very quickly if they detect staging. Applying this environment's statistical standards to that context without unpacking the cultural difference is a mistake. But ignoring data integrity because of cultural difference is a larger mistake.

A cell marked insufficient information is not a failure. It is the boundary between analytical work and imaginative work.

What to track in the coming rounds

If you follow athletics in the period ahead, these are the signals I will be watching instead of watching the medal table.

First, the return of source data. An analytical sheet is only valuable when the source metadata fields are complete, including publication date, competition venue and measurement readings. At many regional meets, these readings are skipped at the publication stage, and every comparison must then be annotated with assumptions.

Second, the competition-load curve. An athlete who starts at several meets in consecutive weeks inside the six weeks before a major event usually pays the price in the semi-final or the final. This is the easiest indicator to measure and the least used.

Third, the wind and altitude baseline of meets inside the points system. When a meet moves venue, the relative value of the entire qualifying points pool changes with it.

Fourth, the distribution of injury data. Until federations publish training-load data, most injury predictions remain guesses dressed in scientific clothing.

I do not believe any single model is strong enough to ignore those four signals. I only believe that a fully annotated model can be corrected. The capacity for correction is the entire value of this work.

The data sheet from that night is still on my machine. I have not deleted it. It is a reminder that the boundary between analysis and fiction is not the length of an article, but whether the writer dares to leave a cell empty.