Three in Four: FIDE's Survey Has the Number, Not the Method

Last week the story ran on four or five outlets at once: a FIDE survey found that three in four female chess players have experienced gender bias or harassment during their careers. I read it three times, and all three times my eye stopped on the line printed underneath — the website's cookie and privacy notice. This is not a joke. When a report is missing its sample size, its questionnaire, its fielding dates and FIDE's own primary document, what survives is the cookie notice. And when a number arrives without its instrument, the job is not to repeat the number. The job is to ask where the instrument went.

The first clue was a scoresheet nobody thought to keep.

My habit is old. In 2026 I was sent to cover the National Women's Championship, where Anupama Gokhale defended her title before forty spectators and one photographer. I filed nine hundred words with three games annotated. The desk ran a hundred and fifty, cut the games, kept the trophy photograph. That night I began photocopying every scoresheet I could find. Thirty years later, the notebook still knows the sound of that hall — and that same habit is what makes me ask which scoresheet produced a figure of seventy-five per cent.

The survey itself is not a small thing. Chess is a sport whose institutional data has, for decades, meant ratings and titles and nothing else. FIDE has now conceded that an experience exists inside its game which the rating list has never recorded. Published properly, this would have been the first written record of a silence I have been filling in notebooks since 2026, through Bled in 2026 and Tirana in 2026. What was published instead was one sentence, with the machinery behind the sentence withheld.

So I did what I do with scoresheets. I went looking for the sample — how large it was and who was in it; for the exact wording of the questions; and for FIDE's primary document. None of the three was available. What was available was an aggregate percentage, repeated without annotation. Everything that follows is therefore not about the number but about its architecture, and the architecture is the actual story.

The largest technical flaw is the one least noticed. The headline says "gender bias or harassment" — two different things, fused into one figure. Bias can include the small insolence of an organiser assuming a woman cannot read an opening. Harassment can include conduct that is criminal. Fusing them produces a statistic that is strong in a speech and useless in a policy room. An organiser cannot tell from this number whether the problem in front of them is one of culture or one of safety, and those two require entirely different remedies. Culture needs training, a code of conduct and mediation. Safety needs a reporting channel, an investigation and removal. A blended figure opens neither door.

The second problem is arithmetic. The sample size is not disclosed. At roughly four hundred respondents, the confidence interval on a seventy-five per cent estimate is about four percentage points; at a hundred, about eight and a half. The number is directionally robust and its precision unknown. For the reader, that means any downstream gloss on "three in four" is inference dressed as measurement. And a bigger weakness than the sample size is the absence of strata — no split between elite and amateur, over-the-board and online, one continent and another. A policymaker needs a target, and an unstratified number does not aim.

The third problem is selection bias, and it behaves exactly the way newspaper mail-in surveys used to behave. A survey circulated through federations and platforms reaches the engaged: the tournament-active, the online-present. That is precisely the cohort most exposed to both bias and harassment. Self-selected online samples almost always produce prevalence estimates higher than population estimates — which makes the figure not false but specific, describing the population still in the arena rather than the population that left it. Nobody counted the ones who left.

One further element of every serious social survey is missing: the control. Male players were apparently not asked the same question. With a matched baseline, one could isolate how much of chess's mistreatment problem is gendered and how much is a general culture of casual cruelty. Without it, seventy-five per cent is not yet evidence of a distinctively gendered experience; it is a number that opponents will dismiss with the single word "everyone." That is the weakness every controversy reaches for, and the survey left the door open.

Jurisdiction went unmentioned entirely. A large share of harassment in chess now happens online — platform chat, direct messages, social media — where FIDE's authority is thin. If online experience was included, FIDE is publishing a prevalence figure for conduct it cannot police. If it was excluded, the figure is partial. Either way, the reader is entitled to know which world the question was asked about, and the reporting does not say.

Now the part that mattered most to me and got no space at all. In a transfer window people watch contracts and agents. In women's chess, the contract is a public-sector job and the agent is a state cash award. The Tamil Nadu machinery that has produced champions since the Aaron era — the Tal Chess Club lineage, the Velammal school schedule, the state's cash-award notification — is recorded not in press releases but in school registers and state gazettes. The ONGC and Railways payrolls are the actual safety net under careers from Humpy to Harika. When we talk about harassment, we are also talking about a structure in which a woman player cannot easily walk away from a salaried post, and the terms of that post — transfers, leave for tournaments, the cost of making a complaint — sit inside the same economic frame.

So does the coaching room and the arbiters' ladder. Power in chess is not only on the board; it is in the right to a seat in the room — who gets a coach, who gets a second, who gets a place at a national camp, who gets a route to an arbiter's licence. This is where bias is most expensive and least recorded, because it has no scoresheet. I went looking for a result and found a whole life — one in which the largest decision a player makes is often which tournament she decides not to enter.

Measure what can be measured. Count girls' entries at under-11, then count their presence at under-17. The gap between the two is the most honest indicator available, because it counts the people who quit and never filed anything. Withdrawal rates from mixed open sections, continuity of participation across a cycle, error rates under time pressure, drop-out by age band — four datasets FIDE already holds, capable of being placed in a single table. Seventy-five per cent is a sentiment metric, not a performance metric, and policy is built on performance.

The structure of women's chess already says where the loss occurs. Judit Polgar peaked at 2735 in July 2026 and remains the only woman to have entered the world's top ten. Of roughly nineteen hundred grandmasters today, about forty are women — under two per cent. That gap is not a gap in talent; it is an accounting of the years girls lose to schooling, travel, safety and permission. When a sixteen-year-old decides she will not enter the open section, she is not deciding she is weaker. She is deciding her time is more expensive than the tournament.

Bled is not a footnote; it is a fifteen-year-old changing the record. I paid my own way there in 2026 because no desk in Delhi thought it was news. What I learned there was technical: I had been quoting results I could not read. Six months later I paid a coach for notation lessons. In Tirana in 2026 I was the only Indian print journalist present as Hou Yifan beat Humpy 5.5–2.5, filing eighteen hundred words a day while Indian papers ran four paragraphs of agency copy. Afterwards I drafted a book; eleven publishers passed, one explaining that it was "a Humpy story, not a chess story." The rejection letters still sit at the front of the manuscript, because they became its architecture.

And then the moment the ledger is kept for. At Budapest 2026 India's women won the Olympiad gold for the first time — Harika, Humpy, R Vaishali, Divya Deshmukh, Vantika Agrawal. I followed those boards through the night, and I will say without hedging that the gold came not from a miracle but from the same quiet machinery Tamil Nadu has run for four decades: the Velammal timetable, the Tal club, the state award, the PSU payroll. That is accounting, not sentiment. A debt was partly paid.

Which brings the contrarian angle I find least comfortable. A seventy-five per cent headline, published without strata and without the instrument, damages the purpose it was meant to serve. A number that cannot show its method will be killed by the sentence every institution loves best — "more study is needed." In practice, more study means one more seminar, one more committee, one more promise. Meanwhile the question that matters goes unanswered: which organiser faced action, whose arbiter licence was suspended, which coach's investigation was completed. The survey named nobody because it began with no case. The institution did not gather evidence; it took the temperature of a mood.

The biggest story here is not the one in the headline. It is that an international federation surveyed its own sport and did not publish the questionnaire. An institution doing accountability shows its instrument first so that someone can audit it. An institution doing communications shows only the finding. I have known members of FIDE's women's chess commission and office-bearers of state associations for years; some will dislike this piece, and I would rather disclose that I treat them as sources, not as exemptions from criticism.

So what now? Budapest gold does not prove the problem is solved; it proves that when the structure is provided, results arrive — which makes the question not whether girls can play, but how many years they will be allowed to keep playing. For anyone who wants to take this figure seriously, there is one practical demand: FIDE should publish the sample, the questionnaire and the raw responses, with fielding dates, so that an outsider can count them again. The day that document arrives, we will stop talking about harassment as an atmosphere and start keeping it as a ledger. Until it does, the sentence about three in four will remain a weight on every girl filling in her first under-11 entry next season — a girl nobody has yet asked whether she will still be playing afterwards.