FIDE survey: three in four female chess players report gender bias or harassment — but the arithmetic behind the number is still missing

The number arrived in a single line. Three in four female chess players around the world have said they have faced gender bias or harassment. Below the line there was nothing. No sample size, no fielding date, not one question from the questionnaire, no link to FIDE's primary document. The page that carried the sentence to me also carried a cookie notice, which told me more about the report than the report did: this was a scraper's copy, not a piece of desk journalism.

I keep a workbook the way other people keep a diary. That night I wrote seventy-five per cent in the left margin and left the right-hand column blank, because a percentage without a denominator is a rumour wearing a lab coat.

This piece is an attempt to fill that column — not by arguing the number is false, but by insisting that the arithmetic which should have travelled with it has not been published. Until it is, the chess world cannot extract from this statistic the one thing it needs: a policy.

Who commissioned the survey, and what we do not know about it

The fact is small and should not be inflated. FIDE, the international chess federation, commissioned a survey of female players, and the finding was that three out of four respondents reported gender bias or harassment. That is all. No player is named. No incident is described. No tournament, no national federation, no date, no sample size.

That is the single most consequential limitation of this story and I would rather state it at the top than bury it.

The instrument itself is not new to chess. FIDE's Women's Chess Commission has carried, for decades, the double duty of designing participation programmes and measuring them. In India the conversation takes a different shape because our pipeline genuinely works — the school-based structures in Tamil Nadu, the steadily lengthening lists from the Velammal network, the AICF's state-wise scholarships and cash-award schedules. Koneru Humpy, Harika Dronavalli, Tania Sachdev, and now R. Vaishali and Divya Deshmukh: a queue whose existence needs no survey to prove. But the distance between climbing onto the pipeline and staying on it is exactly what this survey cannot measure. That distance is probably the important one.

The grid the number sits on

I drew the grid by hand before I ever trusted the software, and in chess the grid is made of numbers, not chalk. That night I built three columns: the top women's rating, the gap to the world number one, and the share of full grandmaster titles held by women.

The structural picture looks roughly like this, and I am flagging clearly that these figures are approximate, date-sensitive and pending independent verification: the leading women sit somewhere in the 2550-2600 Elo band, the world's top player above 2800, a gap of two to three hundred points that has narrowed for years without closing. Fewer than fifty women have ever held the full grandmaster title, under two per cent of the GM pool.

Now place seventy-five per cent on that grid. The number changes shape. It sits on top of a participation structure in which women remain a minority in almost every open hall — my estimate, stated as an estimate, is ten to fifteen per cent of the field at a typical open event.

In 2026, with the world shut and the Goa bio-bubble empty, I watched FIDE's Online Olympiad on the ChessBase India stream at two in the morning. The servers failed, India and Russia shared the title, and one thing became plain: the chat box was the most active chess room on earth at that moment, and no arbiter on the planet had jurisdiction over a single word of it. That fact belongs next to this survey, because a large share of what happens to women in chess now happens on a screen rather than across a board.

The core analysis: one number, six different questions

The headline fuses two distinct problems into a single percentage. Gender bias is a cultural problem — condescension, assumed incompetence, the tone in a commentator's voice that no rulebook can capture. Harassment is a safety problem — conduct that may be criminal, requiring witnesses, records, and a door you can knock on without fear.

The remedies diverge completely. Bias needs arbiter training, commentary standards, media guidelines, school-level sensitisation. Harassment needs safeguarding officers, a named contact at every event, independent investigation, and visible discipline for officials. A merged percentage cannot tell an organiser whether he has a culture problem or a safety problem. And an organiser who does not know which problem he has will choose the cheapest response available to both — a poster.

Where is the denominator

The sample size is undisclosed. That is not a small omission. If the sample was four hundred, the 95 per cent confidence interval on a 75 per cent estimate is about plus or minus four points. If it was one hundred, the interval widens to roughly eight and a half. The same survey produces 'three in four' in one case and 'somewhere between two-thirds and five-sixths' in the other.

Journalism can live with that difference. Policy cannot. Policy does not ask whether the problem is large — that is settled — it asks where to place budget, training hours and safeguarding officers, by state, by age band, by level.

One hundred and sixty-nine goals later, the pattern was still hiding, and it only emerged when I split the goals by phase, zone and trigger instead of leaving them in a total. Here I do not even have the goals. I have one percentage. Nobody can act on it unsplit.

Selection bias: the people asked were the people most exposed

Surveys of this kind travel through federation channels and platform mailing lists, which means they systematically reach the most engaged players — the ones who play tournaments, who are present online, who stay plugged into the chess world. That is precisely the group with the highest exposure to both bias and harassment.

This does not make the number wrong. It makes it a measurement of a selected population rather than of the whole. Prevalence estimates from self-selected online samples typically run high relative to population estimates.

There is an inverse distortion worth more attention. The women who have already left chess — and in harassment cases drop-out is the largest invisible dataset of all — are not in the sample, because the sample is drawn from people still inside the sport. What happened to them never entered the percentage.

Seventy-five per cent is a floor, not a ceiling

In small, tightly networked professional communities, the most severe conduct is the least reported. The reason is structural: the player filing a complaint is judged by the same people who control next season's invitations, wild cards, coaching camps and federation appointments. Arbiters, organisers, coaches and officials sit in the same small room. In that room, silence becomes a career strategy.

So the figure should be read as a lower bound. What surfaced is at least this much; what stayed down is probably heavier.

There is a further layer. Bias is an experience not every respondent recognises as bias. A girl raised on the sentence that girls do not last in chess will not necessarily tick that box, because the sentence never sounded like an insult — it sounded like weather. Normalisation is a survey's worst enemy, and chess has been normalising for a very long time.

Jurisdiction: FIDE is publishing data on conduct it cannot govern

This is my sharpest doubt. Harassment now largely happens online — platform chat, direct messages, social media, the comment section under a stream. FIDE governs over-the-board events and the fair-play rules of its own online competitions. It has no writ over a player's private inbox.

If a large share of the survey's responses describe online experience, FIDE is publishing data on a problem its own rulebook does not reach. That is not hypocrisy; it is honesty about the limits of the institution's reach. But the reporting should say so, otherwise the reader concludes the problem lives in the tournament hall and dies with a change to hall regulations.

And there is a conflict nobody discusses: FIDE is simultaneously the body that commissioned the survey, the body that appoints arbiters, and the body that receives complaints. When the complaint channel and the investigating authority are the same institution, reporting rates fall rather than rise. That too is part of the structure.

The performance channel nobody is measuring

This statistic describes no game, no opening, no endgame. It describes experience. There is a path from experience to performance, and it is measurable — it simply has not been measured.

Persistent bias degrades preparation. Concentration falls under time pressure. Most of all, event selection shifts — invitations declined, cities avoided, mixed open sections quietly withdrawn from, an absence that never appears in the rating list because ratings do not measure absence.

The proxies are all available and all missing: withdrawal rates by gender, participation continuity across a cycle, drop-out by age band, error rates in time trouble. Until those four columns are filled, seventy-five per cent measures sentiment, not performance. Policy built on sentiment becomes a speech.

The contrarian angle: the number opened a door; regulations decide what is behind it

Let me first state the conventional claim at full strength, because dismantling a weakened version of it would be dishonest. Chess has a systemic problem with how it treats women, the problem is decades old, and this survey puts a rate on it for the first time. Those who insisted these were isolated anecdotes now face a figure. That is a real gain and should not be minimised.

Now the contrarian reading, quietly. Numbers change minds; they do not change structures. Most of what Indian chess achieved over two decades came not from statistical advocacy but from administrative paper — the PSU appointment lists, Tamil Nadu's cash-award schedules, the school agreements. The same logic applies here. Seventy-five per cent is a date on which a committee showed a slide. The change is the boring artefact that follows: a safeguarding clause in tournament regulations, and a named contact whose phone number is printed beside the players' list.

Second, the political fate of an unmethodical statistic. A number that arrives without a method hands its own rebuttal to its opponents. The sceptical reader will ask for the sample, the date, the questionnaire — and he will be technically correct while being morally wrong. Publishing the method is not academic etiquette; it is the cheapest available defence of the finding.

Third, and least comfortable. I have always treated the separate women's section as a ladder, because it is one: it supplies a first event, a first trophy, a first sense of belonging. But the separation hides a defect. Where the mixed open section offers no daily rehearsal of the question 'who speaks up when someone says the wrong thing', the safe environment becomes a wall around a separate room. Safety is built by habit, and habit is built in crowded mixed halls, not in empty protected ones.

And one last omission. The survey did not ask male players whether they had experienced harassment in chess. Without that comparison question, three in four does not establish that the problem is gender-specific — only that it is very large. I happen to believe the behavioural gap is wide enough that the comparison should be unnecessary. Still, a question left out is always an answer someone else will fill in for you.

What to watch next

The right-hand column of my workbook is still empty, and only one thing will fill it: FIDE's primary document — sample size, fielding period, the full questionnaire, and the stratified tables. When it arrives, three questions should be asked. Are online and over-the-board experiences recorded separately? Are results broken down by respondents' age and level? And most importantly — is there a policy annex, or only a press note?

The next time an organiser in Delhi prints a poster for a rapid event, I will read it closely. Prize fund, age categories, time control — and whether a safeguarding officer's name and number appear anywhere on it. If they do not, that organiser will never read the three-in-four figure; he will treat it as a number hanging in the air. I have seen an empty stadium. My notebook's grid was never empty. Here the grid is empty, and that is the story.