Nine Empty Fields: How Athletics Prices an Athlete When the Data Is Missing

Nine Empty Fields: How Athletics Prices an Athlete When the Data Is Missing

21:40 on a rainy-season Tuesday in Osaka. On the screen in front of me sits a spreadsheet with sixty columns. Forty-three cells contain a single character: N/A. Wind, altitude, shoe model, sole thickness, ratification status, first-thirty-metre split, last-thirty-metre split. To the right of the spreadsheet, a bookmaker's price board has already formed — smooth, complete, no gaps at all. Somebody has finished pricing an athlete whose dossier is empty.

Thirty years of covering athletics taught me something no classroom did: most of an analyst's work happens in the empty cells, not the filled ones. Numbers never lie; the liar is whoever chooses how to read them. But before anyone chooses how to read, someone chooses which columns exist. That is the entire story of this article.

I once sat at Nagai Stadium in Osaka and watched a sprinter run two hundred metres with a tailwind, watched a beautiful figure appear on the board, and watched someone start typing a celebratory post at that exact moment. None of them checked the wind column. The wind column was right there. They simply never opened it.

Context: how an athletics dossier gets built

A professional athletics dossier has four source layers. The first is World Athletics' official results database, holding marks, window dates and competition conditions. The second is meet organisers' operational data: wind gauges, photo-finish images, split times, sensor feeds. The third is medical and compliance records, held by the Athletics Integrity Unit and national anti-doping organisations. The fourth is field observation by journalists, coaches and officials.

The problem is that these four layers are never published at the same time, in the same format, at the same resolution. Layer one is almost always open. Layer two opens depending on the meet, the organiser, the year. Layer three is almost always closed, opening only when a case is charged. Layer four depends on whether anyone patient enough stood on the inside lane.

The result is a structural paradox: the conclusion is always published first, and the data arrives later, in instalments, in waves of confirmation. In the gap between those two moments, the market — media, sponsorship, betting — has already had to pay. And it pays by inference.

I have worked for a large Osaka betting exchange since 2026. My job is to read the price board and find where it is misreading reality. Every odds movement is a pulse; I only hear it when I put my ear to the ground of the data. Some movements are not a reaction to new information at all. They are a reaction to the absence of new information. Those are the most informative ones.

Based on my experience tracking matches and meets across many seasons, I developed a working habit I still keep: before analysing anything about an athlete, I check the empty cells in that athlete's file first. The empty cells tell me who controls the story.

One example far enough away to be safe. In 2026, when new sports platforms were racing to publish sentimental J-League analysis, I published a study comparing the PPDA metric across eighteen clubs. Shimizu S-Pulse were underperforming their expected goals by 11.3 goals. The media called it bad luck. The dataset called it a structural hole in central midfield. Final standings that season: fourteenth, not the eighth the glowing previews predicted. There was no miracle. Recovery is never a miracle; it is simply the thing you saw in the numbers three months earlier.

I now apply that method to an athletics dossier with nine empty fields. Those nine fields correspond to the nine dimensions any serious athlete file needs. The names and specifics change. The structure does not.

Field one: the mark and its reference value

An athletics mark only has value when it carries at least four variables: wind, altitude, shoe model and track condition, and the meet's baseline expectation.

Wind is the most frequently ignored variable and the easiest to exploit. In sprints and jumps, the legal limit for record recognition is a tailwind of no more than 2.0 metres per second. A hundred-metre mark run into a 2.1 m/s tailwind looks like a leap in class. It is usually a data field pushed just past a threshold.

Altitude works the same way. Tracks in Nairobi, Bogotá and Mexico City hand out free aerodynamic advantage in short events and jumps. How to convert that advantage is still debated among analysts, and the debate itself is why a mark set at altitude always needs a footnote rather than being treated as a sea-level mark.

Shoe model is the newest and noisiest variable. From 31 January 2026, World Athletics capped road shoe sole thickness at 40 millimetres, track spikes at 25 millimetres, allowed only one rigid plate per shoe, and required shoes to be available at retail for at least four months before competition use. Early 2026 was a genuine blank-field crisis: leading athletes preparing for the Tokyo Games tested prototype versions, and no column in any international spreadsheet recorded which version they wore, on which day.

The practical consequence: reading a road result from 2026 to 2026, an analyst must ask about shoe model and sole thickness and accept that the answer may not exist. That is the cleanest illustration of the principle — a recorded mark can still be a mispriced mark.

Finally, surface and climate. Tokyo 2026 unfolded in extreme heat and humidity in the Japanese capital, and organisers moved the distance events to Sapporo to reduce heat risk. A marathon run in Sapporo in August cannot be compared directly with a marathon run in Berlin in September. Both are marathons. Both count. They are not the same measurement.

Field two: athlete condition and position on the age curve

A serious athlete file needs four curves: personal-best progression, current-season form, injury risk, and planned peaking.

Athletics has an advantage over most sports: personal-best data is public, dated, geolocated and wind-annotated. Anyone who can open a results page can redraw the progression curve. Precisely because it is so easy to redraw, people forget to read it.

The progression curve says three things. First, rate of improvement. An athlete who improves steadily across four consecutive seasons has a training programme with headroom. An athlete who stalls for three seasons and then jumps has two possible explanations: a major system change, or an external force. Both need checking, and both get skipped because a big jump is easier to write about than a small one.

Second, distribution across events. An athlete drifting from fifteen hundred metres to five thousand metres usually has a specific physiological reason, and that reason determines whether they retain upside or are being pushed out of their best event. The event-by-event record makes that drift visible.

Third, gaps. A long gap in the competition calendar is the heaviest blank field in any dossier. It could be injury, maternity, a suspended sanction, or a strategic decision. Four causes, four entirely different forecasts for the following season, and if the file cannot fill that cell, every automated forecast model produces a large error.

On injury risk, I always separate age from competition load. A twenty-four-year-old distance runner and a thirty-four-year-old distance runner with the same personal best carry different risk, and both differ from a sprinter of the same age. This sounds obvious. It rarely appears on a prediction board.

On peaking: peaking is a planned variable, not a random state. Where periodisation is mature, an athlete hitting a season best immediately before a major championship is the output of a plan. Where that structure is looser, the same phenomenon may be luck. Treating those two cases identically is a serious error, and it is only avoidable if the file includes a training-log column.

Kelvin Kiptum is the most painful example of the limits of any age-curve model. He ran 2:00:35 in Chicago on 8 October 2026, breaking the world record, and died in a road accident in Kenya on 11 February 2026 at twenty-four. Every forecast curve built beforehand became meaningless, which is a reminder that models cannot price finitude.

Eliud Kipchoge teaches a different lesson. On 12 October 2026 in Vienna he ran a marathon under two hours in a privately organised project with pacers and vehicles. That result is not recognised as a world record. On the official database, that run holds no record. On the news feed, it holds one. The gap between those two boards is the whole problem.

Field three: competition structure and qualification mechanics

This is the most expensive blank field, because it is not about how fast someone runs but whether they are allowed to run at all.

World Athletics operates a dual qualification system for major championships: entry standard, or world ranking points. The world ranking system has been in force since 2026, built on the average of an athlete's best results across a rolling twelve-month window, combining placing points and performance points, weighted by meet category.

This creates a reality many fans miss: an athlete can hold a personal best good enough on paper and still not compete, because the standard must be achieved inside the defined window. That window typically runs about a year before the championships, and longer for the marathon to match the two-races-a-year rhythm of most distance runners.

For the Paris 2026 Olympics, published entry standards included 10.00 seconds for the men's 100m, 11.07 for the women's 100m, 48.70 for the men's 400m hurdles, 54.85 for the women's 400m hurdles, 5.82 metres for the men's pole vault, 2:08:10 for the men's marathon and 2:26:50 for the women's marathon. Anyone who has followed a qualification cycle sees the problem immediately: the women's marathon standard is within reach of hundreds of athletes, while each event fields only around eighty to a hundred competitors, depending on the national Olympic committee.

So most tickets do not travel by the standard route. They travel by world ranking and by national quota. And here the blank field appears: national selection mechanisms are not obliged to be published. Some federations publish criteria before the season. Some publish after. Some publish nothing.

For a data analyst this is a partially quantifiable blank, obtained by comparing three lists: the entry standards, the world rankings at the cut-off, and the final entry list. Those three always diverge, and the divergence is where the story actually lives.

Competition strategy is another overlooked variable. Meet density inside a qualification window is a cost decision. Racing often to accumulate ranking points means accumulating fatigue, and at middle and long distances that fatigue does not show in the next race. It shows in the most important one.

Sifan Hassan is the clearest case of a strategy that models bet against. At the Tokyo 2026 Olympics she won the five thousand metres, won the ten thousand metres and took bronze at fifteen hundred metres. At the Paris 2026 Olympics she entered three distance events and won the marathon in 2:22:55, having already raced the five thousand and ten thousand metres earlier in the same Games. Any simple physical-cost model puts her outside marathon medals. She won.

Ethiopia's Paris 2026 strategy shows the value of a blank filled on time. Tamirat Tola, not on the original list, was substituted for Sisay Lemma and won the men's marathon in 2:06:26, a course record. That substitution was not luck. It was the output of a reserve system maintained throughout the cycle.

Here sits a lesson about record categories. Peres Jepchirchir won the London Marathon on 21 April 2026 in 2:16:16, a women-only world record. World Athletics distinguishes mixed-race and women-only records, and that distinction completely changes the meaning of an identical time. Anyone reading a results list without reading the category will always misread the value.

Field four: event landscape and national strength

Discussions of national strength in athletics drift easily into cultural description. I do not read the world that way. I read three indicators: depth of the elite group, talent pipeline, and the rate at which personnel circulate between training centres.

At long distances, Kenya and Ethiopia hold position through depth rather than individuals. In the men's and women's marathon, the number of athletes from these two countries running faster than the Olympic standard consistently exceeds the number of quota slots each country may use. This is a distinctive condition: surplus labour creates an internal market where championship slots are contested domestically before they are contested internationally.

In sprints, the United States holds the centre, and its selection mechanism shows it: a single national trials meet decides the team regardless of world ranking. This is a hard mechanism. It produces shocking omissions and it hands championship places to young athletes who peak at the right moment. It also turns a national meet into an event with higher predictive value than most international meets.

Jamaica deserves analysis because its pipeline is built on schools and a national high-school championship. That is a model of extremely high competitive density at very young ages, and it explains how a country with a far smaller population has sustained a sprint talent flow across decades.

Japan is the case I track most closely, because I live and work here. The university ekiden system, centred on a two-day event in early January, produces a dense talent pipeline and a large volume of young runners logging high mileage. To compare by measure rather than prejudice, look at two indicators: the number of athletes running faster than championship standards at distance events, and the number of medals at world championships and Olympic Games. The first is very high. The second is much lower.

That gap is a blank field with real analytical value. It suggests the ekiden system optimises for one kind of capacity while the international stage optimises for another. There is no conclusion about people here. Only a conclusion about how system design determines which indicator runs high.

On the European side, the Netherlands emerged as a force at both middle and long distances over the last cycle. What matters for an analyst is that Dutch training centres publish a fair amount of process data. When a country publishes process data, the predictive value of its results rises, because hidden variables become testable rather than merely inferred.

Field five: rules and anti-doping

This is the highest-risk blank field, and the one sports media handles worst.

Professional anti-doping runs on four pillars: whereabouts obligations enabling no-notice testing, the athlete biological passport tracking blood and urine markers over time, the therapeutic use exemption process, and the violation adjudication system. All four share one property: they operate in secrecy until a case resolves.

That creates a systematically asymmetric information structure. Throughout an investigation nobody knows anything. On decision day everybody knows everything. There is no middle spectrum. And across that missing spectrum, marks are still recorded, records are still set, sponsorship contracts are still signed.

For an analyst, this forces one rule: competition results and compliance status are two independent datasets, and joining them is an act of inference, not an act of reading. I do not join them. I note the uncertainty level and move on.

On technical rules, the most overlooked element is record ratification. A performance on a track does not automatically become a record. It passes verification of course measurement, wind-gauge calibration, shoe compliance and doping analysis. That takes time, and during that time the accurate status is pending ratification. Headlines almost always delete that status.

One example of how reframing conditions changes the value of a result: the women's marathon record set in Chicago on 13 October 2026 in 2:09:56 was immediately called a world record in headlines, while on the official database it sat pending. Both descriptions are defensible in a sense, and that ambiguity is exactly what a data reader must handle alone.

Mispronouncing a name is not the error; the failure is not seeing the outline of a system. I say this from personal experience. In June 2026 I was invited as a data commentator for a test broadcast on DAZN Japan for the Japan versus Colombia match at the World Cup in Russia. In the first half I mispronounced midfielder Hotaru Yamaguchi's name three times. I thought that was the fatal error. What kept me awake for weeks was the goal conceded in the thirty-ninth minute. Tracking data showed one team's average unit length stretched to forty-two metres, breaking the pressing structure entirely. I spent a month rewatching every group-stage tape to correct how I read the match.

The lesson I kept: a mispronounced name is a pronunciation error. Not seeing a system breaking is a professional error, and it is far worse, even when nobody notices.

Field six: team and training system

In most athlete files, the coaching column is filled in formally with a name. A coach's name tells you nothing about coaching capability in a specific context.

Three variables belong there instead: fit between training philosophy and the athlete's physiology, technical and rehabilitation support capacity, and the stability of the training group over time.

The third is the easiest to measure and the most ignored. An athlete who stays in one training group for five consecutive years has a continuous data line. An athlete who changes coaches three times in three years has a segmented data line, and every model built on that athlete's past carries systematic error. This is measurable simply by counting group changes over a fixed period.

On periodisation, the thing to check is not whether an athlete trains at altitude but how volume, intensity and density are distributed across weeks. Without that column, every comparison between athletes compares outputs with no inputs.

Here appears the most common data trap in athletics: training marks. This is the most gossiped and least verifiable category of information. A training performance has no officials, no wind gauge, no doping control, no certified track. It is valid as process data. It is not valid as results data. Confusing the two is one of the largest error sources in sports forecasting.

On technology support, I separate three levels. Level one is collecting process data: sensors, lactate measurement, cardiac data. Level two is using that data to adjust the in-season plan. Level three is building individualised forecasting models from multi-season data. Most training groups sit at level one. A few at level two. Very few at level three. That gap is among the most underpriced variables in long-range forecasting.

Field seven: the risk map

A risk map for a track and field athlete has six categories: competitive, doping compliance, financial and career, eligibility and rules, public and brand, and systemic.

Competitive risk is the easiest to estimate because it rests on results data. Financial risk has shifted faster than any other category in the past decade. The income structure of elite athletes has moved away from prize money toward equipment sponsorship, appearance fees and personal contracts. That means personal brand risk has become a core component of career risk, and it appears in no results table.

Systemic risk is the least visible. Record ratification structures, calendars and quota allocations can all change between cycles. A small change in how world ranking points are computed can invert the competition strategy of hundreds of athletes, and it generates no headline compelling enough to read.

During a transfer window, the same risk structure appears in team sports in a different form. There, the blank field is named release clause, instalment structure and performance-based payment conditions. Transfer headlines quote the fee. The contract contains the structure. The release clause and the wage bill are the real story, and they are also the hardest part to verify.

What gets called transfer market noise is usually surface paint over a deeper order: the order of who is permitted to publish what, and when. People rush to read rumours because rumours are published densely and for free. Nobody reads contract documents because they are published sparsely and at a cost. That is the entire mechanism.

Field eight: public narrative and expectations

Every athlete has two files: the data file and the media file. They run on different cycles.

The media cycle has four phases. Discovery, when an unusual result appears. Diffusion, when the story is retold by more channels with decreasing detail and increasing emotion. Consolidation, when the story becomes the default expectation and contradicting it becomes expensive. Correction, when reality forces the story to change.

An analyst works best at the boundary between discovery and diffusion, and worst during consolidation. In consolidation, any analysis running against expectation is read as a hostile statement rather than a calculation.

Testing whether a public narrative is sustainable is straightforward. First, check the sample. A single mark in a single meet is not a stable ability level; it is one data point. You need at least three points across different conditions before discussing an ability level. Second, check the foundation. Is the story supported by a multi-season progression or by one race. Third, check the timeline's feasibility. Expectations of a large jump inside one season usually violate the athlete's own physiological limits.

When everyone looks one direction, I start examining the gap behind their backs. In athletics that gap is usually the competition-conditions column and the ratification-status column. Both sit at the far end of the spreadsheet, and nobody scrolls there.

The most useful indicator in this category is the ratio between how much a result is discussed and how much it is verified. A result mentioned thousands of times with a single originating source has a high ratio. A result mentioned less often but with multiple independent sources — organisers, national federations, the international database — has a low ratio. This indicator sounds dry. It forecasts better than reading commentary.

Field nine: transmission into the wider industry

An athletics result does not stay inside athletics. It travels through six channels: competition commercialisation, equipment technology, representation and sponsorship, the youth talent chain, adjacent markets, and the national team ecosystem.

Equipment technology is the fastest channel and the noisiest. When a new shoe appears at a major meet, transmission follows a fixed order: elite athletes wear it, media note the performance, fans infer the equipment, mass running buys it. The variable lost in that chain is the equipment's actual contribution to the performance. Separating equipment contribution from ability contribution has no standard solution, and anyone claiming one is selling something.

The youth talent chain moves more slowly and matters more. When a country produces a successful generation, youth participation rises over the following three to five years, and that effect takes roughly seven to ten years to become international results. That window is the basis for forecasting the global competitive structure a decade out, and it is almost never used.

Representation and sponsorship is the channel analysts skip because it is hard to quantify. It has one simple indicator: the number of announced personal sponsorship deals divided by months of competition in a season. In some countries that ratio spikes after a big result, and the spike precedes a rise in performance the following season. It measures commercial pressure before that pressure shows up in results.

The national team ecosystem is the most underpriced channel. A country that builds a stable training cycle raises the value of every domestic result, because those results become signals rather than outcomes. A country without such a cycle produces the same result as an outcome only. The difference is not in the results table. It is in the training-log column, and that column is almost always empty.

The contrary view: sometimes the blank cell is the correct answer

Having spent most of this article on the damage blank fields do, I have to state the opposite, because otherwise I am selling a different product: the belief that every blank can be filled with data.

They cannot.

Occam's razor applies to sports analysis with uncomfortable force. When a blank field exists, the simplest explanation for it is usually correct: the meet's publication process is not detailed enough, the organiser never collected that data, the national federation chose not to publish. These three reasons explain most real-world blanks. We tend to skip them in favour of a more attractive hypothesis, usually involving deliberate concealment, because attractive hypotheses are easier to write.

My profession has one occupational temptation I must name clearly: the temptation to appoint myself judge of how others read data. My signature line — the liar is whoever chooses how to read them — easily turns an analyst into a magistrate, when his actual job is auditor. An auditor identifies the discrepancy and leaves the verdict to whoever holds authority. A magistrate rules on everyone's behalf.

So my rule is: always present the opposing reading in its strongest possible form before showing where its method fails. If I can only refute an argument after weakening it, I have refuted nothing.

A second and more serious failure mode: emotional data. Market euphoria after a big result is usually treated as noise, when it is in fact structured raw data. It can be defined, measured and tracked. Three indicators suffice: the diffusion speed of a result over the first seventy-two hours, the ratio of emotional to technical commentary, and the lag between the moment a result occurs and the moment prices adjust.

Treating emotion as raw data is the only way an analyst avoids treating it with contempt. And contempt for emotional writing is a professional failure, not an intellectual posture.

A third failure mode is the instinct to find deep order. My signature line about surface paint is a promise: under the paint there is always an order. That promise holds most of the time and fails in a few important cases, and failure always costs more than success, because it makes an analyst skip a simple explanation while remaining confident that he is seeing deeper than everyone else.

The check I use against myself: if a supposed deep order can generate two opposite forecasts without the model changing, it is not a model. It is storytelling wearing a data costume. An era does not begin with technology; it begins with a question sharp enough to cut through the worn path. The sharp question for a sports data analyst is always: how could this hypothesis be proven false, with which data, over what time horizon. Without those three answers, everything else is literature.

Signals to watch in the next cycle

I close with observable signals. Each has a way to observe it, a trigger condition and an expected impact.

First, how many meets publish split and sensor data to the public. This is the easiest to observe. The trigger is at least one top-category meet publishing data in an open format. The impact: the predictive value of results at that meet rises sharply, because hidden variables fall.

Second, the average time between a major performance and the publication of its ratification status. The trigger is that time stretching beyond the usual. The impact: the gap will be filled by inference, and inference is where value gets mispriced.

Third, transparency of national selection criteria in countries with deep athlete pools. The trigger is two countries in the same tier choosing different mechanisms in the same cycle. The impact: the performance gap between them will be misread as an ability gap when it is a mechanism gap.

Fourth, the ratio of emotional to technical commentary in the first seventy-two hours of a transfer window or a major qualification window. The trigger is that ratio exceeding anything seen in the previous cycle. The impact: a high probability of a sudden price correction once real data is published.

Fifth, how many athletes publish training process data on a regular schedule. The trigger is the emergence of a group of athletes treating this as a competitive norm. The impact, over the long run, is the strongest signal on this list, because it moves data from a verification tool to a forecasting tool.

I keep a sheet with exactly these five indicators and update it weekly. When people ask why I do not write more about the athletes being discussed most, I usually answer with a different question: in that athlete's file, which cell is empty. The answer to that question determines almost the entire value of the next article.