Athletics Data Audit: An Empty Framework, and the Error of Reading 'Not Assessed' as 'No Risk'

For eleven years I have read the athletics record book the way you read a file with pages deliberately torn out. The most dangerous thing in an audit is never a wrong number. It is an empty cell — because you can argue with a wrong number, re-measure it, take it apart. An empty cell gives you nothing to hold.

What surfaced this time was not a wrong number. A nine-dimension athletics analysis framework rendered in full — performance assessment, athlete condition, competition structure and qualification, domestic and global landscape, rules and anti-doping, training systems, risk matrix, public narrative, industry transmission — and every single cell carried the same text: "N/A — insufficient information, cannot assess."

A rendered framework looks complete. That is not the same thing as being filled. That sentence is the whole story.

The thread started with one number I could not verify — because there was no number.

Context: what Stage-1 owes the audit

Every project I take on opens with a data audit: source, timestamp, verification method. Three things. Nothing moves until those three are satisfied. In athletics that discipline is not optional, because a single mark means four different things in four different contexts — wind-aided, altitude-aided, in carbon-plated shoes, on a fast track. Sprint and jump marks count for records only with a tailwind under +2.0 m/s. Venues above roughly 1,000 m help sprints and jumps and penalise endurance marks. Road shoe soles are capped at 40 mm. Since 2026 a single false start means disqualification. Qualification runs on two channels — the direct standard, or World Ranking points. On the anti-doping side sit the Athlete Biological Passport, the whereabouts rule (three missed tests in twelve months is a violation), the DSD regulations, and the ANA route for athletes from suspended federations. All of it lives inside the same framework that came back empty.

Stage-1 was supposed to return: title, source, article type, core viewpoints, information points, entities involved, time sensitivity, and source quality. Eight fields. Eight empty. One field populated — domain label: athletics — and even that arrived with no athletics entity attached to it.

Core: the audit in three stages

Stage one — two hypotheses, one of them implausible. Hypothesis A: the source article genuinely contained no athlete, no competition, no mark, no venue. Hypothesis B: the text-deconstruction pipeline failed at the title/source layer and entity extraction never executed. I discarded A. Any athletics item that clears editorial publication contains at least one of those four elements; all four missing at once is a property of process, not of content. The failure has been located, and it sits in our ingestion chain, not in the source — which changes everything downstream.

Stage two — what the null blocks. I had to write the list out, because "could not be checked" and "checked and cleared" cannot share a cell. No wind reading, so no true-value correction. No altitude figure, so that correction could not run either. No shoe model, so the equipment dividend stayed undeducted. No athlete name, therefore no date of birth, therefore no age curve (sprint peak roughly 24–29, distance 26–31, throws 28–33, marathon past 35). No year-by-year PB series, so the anomaly test — a single-season gain beyond roughly three times an athlete's historical annual average — could not be run. No injury history, so no glass-body grading. No competition named, so no tier classification: Tier-1 Olympics and World Championships, Tier-2 Diamond League and continental meets, Tier-3 Continental Tour and national meets. No nationality supplied, so it is not even established whether a US-style trials system applies — the structure that can eliminate a reigning world record holder in one race.

Not one of those cells can be read as clean. Every one of them is unavailable, and the distance between those two words is the subject of this report.

Stage three — three failure modes, and the third is the most expensive. Mode one: a wrong number. At the 2026 Federation Cup men's 400 m final, three manual timings sat 0.3 to 0.5 seconds outside the photo finish; the report on it was titled "The 0.3-Second Lie" for a reason. Mode two: a duplicated number. On the data desk for the Paris 2026 men's javelin final I watched the live feed carry Neeraj Chopra's third throw as 89.45 m twice, and the graphics repeated it; cross-checking the Omega system against the AFI manual log confirmed 89.45 m, and Arshad Nadeem's 92.97 m was flagged as the Olympic record. Mode three is what arrived this time: no number at all. The first two show themselves. The third shows nothing except a document that looks finished.

My three-source rule — primary sensor, official log, manual backup — answers two of those three modes. It had nothing for the fourth case, where there is no source to check. It does now.

Stage four — the largest risk is not inside the framework, it is in the reader of the framework. The report recorded its overall risk rating as N/A, correctly. But N/A does not mean low risk. The anti-doping checklist returned no red flags because no testing history, no ABP status, no whereabouts record was ever supplied. Absence of evidence is not evidence of compliance — an old sentence, and the one most reliably forgotten. Retrospective-testing exposure, the ten-year sample storage and medal reallocation mechanism, could not be evaluated for anyone, because there was no anyone.

The one piece of good news is procedural: the framework is already standing. Restore the source text and it can be filled without rework. I rebuilt the whole audit from scratch, then checked it twice, and the finding held — the work is not incomplete, the input is.

Contrarian: I did not sit down to fill the gaps

Counter-intuitive discovery is my habit, so the temptation arrived on schedule: find a discrepancy, justify the piece. I did not. Where an accepted number survives scrutiny, saying so plainly is a finding; where there is no number at all, the honest move is one paragraph of open admission. The real risk is that a rendered template reads as finished — and the industry pressure to fill cells is strong enough that the next analyst will plug the blanks with inference, because a blank page looks like failure and a page of N/A looks like a product.

The fault is systemic, and it should be named that way. Athletics has no dedicated beat. Most outlets carry a handful of bylines in a non-Games year and dozens in an Olympic year. Wire copy sticks to wire copy. The 0.01 seconds at Los Angeles 2026 survived for decades because nobody picked up the primary source. This audit is the next link in that chain — except this time the inheritance was not a wrong figure but a missing one.

Takeaway: four signals to track

First, source recoverability — re-ingest the original document and a title plus at least one information point should return. Second, extraction integrity — information-point count against article length should never be zero. Third, domain-label reliability — an "athletics" label with no athletics entity is itself suspect. Fourth, time-sensitivity backfill — without a publication date, neither a peaking window nor a news cycle can be read.

The question is not that one report came back empty. The question is how many athletics reports went out this week carrying a cell nobody will ever open, because the form at the bottom looked fine.