Three in Four Female Chess Players Report Bias or Harassment — The FIDE Survey and Its Empty Spaces

Mahabalipuram, July 2026. Eighth day of the 44th Chess Olympiad, the afternoon cafeteria break. In the corridor a woman player caught my wrist and said, “Switch the recorder off.” I switched it off. Then she talked — about group chats, about a coach, about a name she did not want said aloud, and about the question she would have asked at the press conference if there had been fewer cameras in the room. I filled two notebooks across those twelve days, one for moves and one for overheard sentences. That evening the second notebook got a single line: some data never arrives in an interview; it arrives holding your wrist in a corridor.

A FIDE survey now puts a headline on that corridor: three in four female chess players report having faced gender bias or harassment. The corridor has become a percentage. My first reaction was not that the number looked inflated. It was that it looked low — and that the reason it looks low is not the survey but the reporting of it, which never tells us who was asked.

What is known, and what is not

The public information is thin. FIDE commissioned a survey; three in four respondents among female players reported bias or harassment. Everything that would make the figure usable is missing from the secondary report I have seen: sample size, countries covered, fielding period, the questions themselves, the answer options, whether the instrument was online or administered at events, and whether male respondents were asked an equivalent question. There is no link to the primary FIDE document. I am citing the figure because its direction matters, and I am recording alongside it that verification against FIDE's official publication is still pending.

That gap is not small. “Have you ever been underestimated” and “has someone with power over your career asked you something he had no right to ask” are different questions. Fused into one column they make a sharp headline and a blunt policy instrument.

Bias and harassment: one percentage, two diseases

Bias is what makes a 2,400-rated player hear, through her father, that she plays nicely. Bias is the joke that lands in the arbiters' room, the commentary that invents a phrase like “feminine moves,” the training-camp advice that it is time to focus on studies. Harassment is a different category — the 2 a.m. message, the sponsor who wants something settled off the board, the warning that a complaint will cost you the next invitation.

Collapsing them into a single 75% means an institution cannot tell whether it has a culture problem or a safety problem. The first is fixed through media, language, training and the behaviour of officials — slow, quiet work. The second needs a reporting channel, independent investigation, immediate suspension and criminal process — fast, legal, uncomfortable. One number lets an organisation claim it is doing both. That is the survey's biggest strategic weakness.

The architecture of power that no survey measures

Chess is one of the few sports where a single man can be a child's coach, mentor, talent manager and moral guardian at once. In India the intensity is greater because family savings, the child's age and the coach's control sit on the same axis. The coach knows how much the family has borrowed, which tournament is being played on which hope, and whose phone call produces the next invitation. That knowledge is power, and the abuse of it does not appear among anyone's answer options.

Then the arbiter, whose ruling inside the hall behaves like a final one, because an appeal is expensive and paid for in reputation. Then the organiser, holding a short list of closed invitationals, sponsor events and appearance fees. Then the federation, holding selection, camp allotments and clearance to travel. At all four stops the complainant may end up standing in front of the same person she has complained about. The strongest indicator of harassment in chess is not how many incidents occur but how long the route to a filed complaint is.

Seventy-five per cent of whom?

The technical problem is not sample size. It is the absence of strata. If the instrument was distributed online through federations and platforms, it over-samples the most engaged, tournament-active and online-present players — exactly the group most exposed to both bias and harassment. Self-selected online samples routinely produce prevalence estimates above population estimates. That does not make the figure wrong; it defines its limits.

The arithmetic is plain. At n≈400, the 95% confidence interval on a 75% estimate runs roughly ±4 percentage points; at n≈100 it runs ±8.5. The direction is robust, the precision unknown. My real objection is what the 75% refuses to divide — elite from school-level, over-the-board from online-only, one region from another. A survey that cannot say whose 75% it is cannot tell policy where to aim.

The measurable thing nobody measured

In chess the damage of bias rarely shows on a scoresheet. It shows on a calendar. A 2,350-rated player does not stop playing; she narrows. She drops mixed opens, avoids certain cities, quietly declines certain organisers, moves online, and eventually appears only in women's sections. Her career does not collapse, it contracts. Contraction leaves no trace in a rating list, because a rating list records attendance, never absence.

The proxies are cheap and available: invitation acceptance rates by gender, continuity of participation across a cycle, drop-out rates between 16 and 22, error rates under time pressure, and most useful of all, which organisers show a falling share of female entries year on year. Those are the numbers that would convert 75% from sentiment into policy. Nobody is publishing them.

The other board, the one off the table

In August 2026 I live-blogged an online final for eleven hours while six boards froze in sequence. At 3 a.m. the screen showed a spinning circle, and the actual event was a room full of stunned people. The loudest thing that night was the chat going quiet. A large share of harassment in chess now happens in chat windows, direct messages and social media — territory where FIDE's authority is largely notional. A screenshot becomes evidence, and the same group chat functions as both support network and threat.

An honest disclosure here. I filed about 9,000 words from that 2026 Olympiad; roughly 1,800 ran. What was cut were the names that came off the record. Indian chess solved reach around 2026 — the world's largest chess media operation is Indian and its live coverage travels in millions. What it has not solved is documentation: a record with names, dates and outcomes. When the long record is missing, the strongest evidence left is the weakest kind — a percentage. The 75% echoes precisely because the space is empty.

The old line: “the board doesn't know you're a woman”

The board has no senses. The question is who reaches the board, and the board does not decide that. Our collective memory is built out of coronations — Tehran in 2026, Mexico City in 2026, Moscow in 2026, and last December in Singapore, where an eighteen-year-old became the youngest undisputed world champion. We will repeat those dates for five generations. What we will not repeat is the semifinal that was never played, the invitation quietly deferred, the player who left at nineteen and exists now in no file beyond a rating list.

The second unease is about how the survey itself will be used. Institutions turn numbers into posters faster than into policy. Announcing a percentage is far cheaper than naming a case and suspending an official. The risk is real: this 75% will be quoted for a decade as though it were fielded yesterday, while behaviour and reporting norms both drift. A static number hardens from fact into folklore.

What I am waiting for

Three things before I quote this figure again. The methodology — sample, strata, questions, dates, and an equivalent baseline question for male respondents, so the gender-specific effect can be isolated from the ordinary rudeness of a tournament hall. One named consequence — a safeguarding officer at every federation-affiliated event, with a complaint route that leaves the organiser's office. And the withdrawal data that comes out of calendars rather than quotations.

At 1 a.m. the arena was a phone screen and a prayer; I learned that at seventeen, watching a buffering stream from Riyadh. On the same screen now, a message gets typed and deleted, because the woman typing it knows her invitation could be cancelled by morning. The number has arrived. The place to file a complaint has not.