The Empty Object: The Blank Space in Women's Athletics Data

On Thursday night I printed a fourteen-page file at the office. Nine sections, full tables, bold headings. In every content cell, the same line: "N/A — insufficient information." The first page stated that the athletics article just processed by the extraction system had returned zero information points. No headline. No source. No athlete's name. Not one lap, not one competition date. The only field to survive the entire pipeline was a label: athletics.

I read it a second time and realised I was holding an article about the absence of an article. The report calls the phenomenon by a technical name: a silent pipeline failure. The output object still conforms to the schema, still passes structural validation, but carries nothing inside. It raises no error. It simply says, quietly, that there is nothing to say.

In my trade, this kind of break is harder to handle than a red alert. A red alert forces someone to fix something. An empty object slips through the door, settles into the archive, and three months later somebody opens it, sees a full set of blank spaces, and fills them in from memory or from guesswork. The report identifies exactly that point: it rates the risk of downstream fabrication as the highest category, above every sporting risk it was meant to assess. The reason is simple — it has not assessed any sporting risk at all.

The three diagnostic branches offered for the failure are worth reading too. One: no article body was ever passed to the extractor. Two: there was an article, but the analysis layer returned a default object. Three: the source had nothing to do with athletics, and the "athletics" label is a stale value left behind. All three lead to the same place — nothing can be concluded about the content, only about the state of the information.

The part I lingered over longest is where the report blocks itself. It lists five minimum requirements for a re-run: headline, source, publication date; at least three information points, one of them quantitative; named entities; provenance for any mark, including wind reading, altitude and competition shoes; and an assessment of source reliability. Then it states plainly that the document must not be cited or summarised as an assessment of any athlete, event or organisation. A document that gags itself. I rarely see that.

The empty object is old. It is the familiar shape of the women's athletics archive. In 2026 in Amsterdam, women ran the 800m at an Olympic Games for the first time. The event was then removed from the programme and did not return until Rome 2026. Thirty-two years. Across those thirty-two years no results table was left blank, because no table had ever been created.

There are contests whose score is never recorded, because people chose to forget.

The women's marathon entered the Games in 2026 in Los Angeles. The women's 5000m in 2026 in Atlanta. The women's 3000m steeplechase in 2026 in Beijing. Women's ski jumping in 2026 in Sochi. Kathrine Switzer ran Boston in 2026 with a bib number, and only in 2026 did the organisers formally open the race to women. Each marker on that list is a long stretch of time in which women's performances existed on the track but not in the record.

Based on my experience covering matches, the blank space is not where the technical difficulty lies. In 2026 I went to watch Boston Breakers play in the NWSL and interviewed Morgan Andrews, then a reserve forward. She talked about waiting shifts to cover travel costs. That story I still have. What I do not have is the club's full season data set: in January 2026 the Breakers ceased operations, the NWSL went into the following season with nine teams, and part of the club's record vanished from the usual lookup pages.

There are playing lives that never appear on a transfer list, only in the silence between two touchlines.

Transfers have seasons, but the memory of a women's player does not follow any transfer window.

Women's results survive in very modest ways. A referee's handwritten sheet. A clipped local newspaper page. A coach's spreadsheet emailed in 2026. I have a habit of photographing results boards at every meeting I attend, including meetings with four athletes on the 400m start line. Not because I like photographs. Because that photograph is often the only copy left after the official results page is taken down.

Reading that fourteen-page file as a data set rather than a broken report yields a figure worth noting. Nine dimensions, roughly forty content cells, exactly one with a real value: the domain label. The null rate is about 97%. The report also notes that even that label may be wrong — it could be a system default rather than a classification result.

An empty object, read correctly, is more honest than an object filled in by guesswork. That is the entire value of the paper I printed on Thursday night. It does not lie about what it does not know.

A system that returns an empty object for an athletics article is a miniature copy of a larger system: the one that decides which events get timing chips, which meetings get live results, which clubs get an archivist, which seasons count as worth recording. That pipeline is not broken. It runs very well, exactly as designed, and it returns blank space systematically.

The blank space in women's athletics data is rarely a technical accident; it is a decision repeated across decades, changing only its form. In 2026 the decision sat in a committee room. In 2026 it sits in a configuration file nobody reads.

I write to recover the names that history edited out of the line-up.

The report recommends blocking publication. I agree, for a different reason. People tend to fear the empty object because it looks like failure. To me it is the most honest document women's athletics produced this week. What is more frightening is the filled object: a profile written from a press release, a "new Marta" headline built on three matches, a handsome statistics table with no wind provenance. Those cannot be re-run. An empty object can, and it costs one fetch cycle.

This industry measures output in articles, not in verified information points. A pipeline optimised for populated fields will always find a way to populate them. And the way it populates is usually the old way: a good story, a pretty name, a statistic nobody checks twice. I saw it at a sports-data conference once: thirty people talking about how to collect, nobody talking about how to keep.

If the gate the report proposes — headline non-null, at least one information point — were fitted to the whole archive of women's athletics, how many seasons would be returned to an empty state? How many names would drop out of the line-up simply because no line of data stands behind them?

I still keep the fourteen-page file in a drawer, next to the notebook recording the times I went to a venue only to ask one person about a match nobody entered into the results table. That gate, if it were ever installed, would not save anyone. It would only make sure that next time we know precisely what we lost.