FIDE Survey: Three in Four Female Chess Players Faced Bias or Harassment — But Who Audited the 75%?

Kolkata, November 2026. The first edition of Tata Steel Chess India. I took the creator pass, skipped the press room, and sat in the third row of the rapid and blitz arena counting heads for three days. The hall was half full. Viswanathan Anand and Hikaru Nakamura were playing 15+10 on the board while the broadcast treated silence as reverence. That evening I wrote that Kolkata has fourteen million people and this tournament does not have a single street poster. Kolkata didn't shake hands. It sat and watched, and the camera mistook that for silence.

Seven years later, another number arrived. A FIDE survey reports that three out of four female chess players have faced gender bias or harassment in their careers.

I am not calling that number a lie. I am saying that in a season when chess's biggest money is signing contracts, we are waving a percentage whose denominator, sample size and fielding date nobody has bothered to ask for. If I were running this as a match thread, my first tweet would be simple: big number, empty back end.

Context: the story is not the statistic, it is how fast the statistic travelled unexamined

What we have is short. FIDE commissioned a survey. Female chess players were asked whether they had encountered gender bias or harassment. The result: three in four, roughly 75 percent.

What we do not have is a long list, and every missing item sits exactly where policy gets made. No sample size. No indication of which countries or regions were covered. No fielding period. No distribution method, no response rate. No link to the primary FIDE document. No named player, no named tournament, no named national federation.

The material this report is built on also contained a website's privacy and cookie notice among its key information points. That reads as a technical detail. It is not. It means the source page was scraped by a machine rather than read by a person. The headline figure does not become false because of that, but confidence drops that whatever was dropped was irrelevant. The most usable information — what FIDE plans to do next, which panel gets formed, whose budget pays — usually sits in precisely the section that condensed secondary reporting cuts.

One structural benchmark matters here, because the number did not arrive in a vacuum. The world number one in the open section sits above 2830; the highest-rated woman has lived in the 2550–2620 band, a gap of roughly 230 to 280 Elo that is narrowing slowly (figures approximate, pending verification). Fewer than 50 women have ever held the full grandmaster title, under two percent of the entire GM pool. That is the shape of the room in which the complaints are filed — small, dense, where every organiser knows every player by name.

Core analysis: one number, two different problems

The question asked was "bias or harassment." A disjunctive question. The answers were melted into a single percentage and turned into a headline.

That is the widest crack. Bias is a culture problem: a woman being told she arrived through a quota, a commentator calling her preparation a surprise, an opponent visibly irritated at sharing an open section. The remedy is training, a code of conduct, and women on commentary panels. Harassment is a different category, containing conduct that is criminal in many jurisdictions. The remedy is a complaints mechanism, independent investigation, physical safety, suspension.

An organiser has a right to know whether it faces a culture problem or a safety problem. The 75 percent figure cannot tell them, because both questions were dropped into the same box. That number is magnificent in a keynote and useless at a budget meeting.

Nobody is asking how big the sample was

The arithmetic is simple and needs no specialist. If roughly 400 women responded, the 95 percent confidence interval around a 75 percent estimate is about plus or minus 4 percentage points. If the respondent pool was closer to 100, that interval widens to about plus or minus 8.5 points. The true figure could be 66 percent or 83 percent.

The direction is clear — a figure that large does not appear by accident. The precision is unknown, and policy is built on precision, not direction.

The larger danger is selection bias. Surveys distributed through federation emails, platform notifications and player groups reach the most engaged, tournament-active, online-present players — precisely the group most exposed to both bias and harassment. Prevalence estimates from self-selected online samples routinely run high relative to population estimates. That does not make 75 percent false. It makes 75 percent a signal, not a census.

The second absence stings more. There appears to be no baseline question put to male players — have you experienced harassment in chess? Without a comparison group we cannot separate a gender-specific effect from the general culture of the sport. Without a control line, every bias statistic stays half-finished.

The missing strata matter more than the missing sample size

Suppose the sample was 400 and the error only plus or minus 4. The 75 percent is still an average that seals an elite grandmaster and a seventeen-year-old district player into one bag. It does not separate over-the-board from online. It does not separate Kerala from Tamil Nadu, Maharashtra from Delhi.

A policymaker needs to know which cohort carries the risk. If harassment is mostly happening in platform chat and direct messages, what use is a district association's conduct rule? If it is mostly happening in coaching camps and tournament hotels, what use is a platform moderation policy? An unstratified number makes noise. It does not aim.

And the most serious conduct is the least reported

In small, tightly networked professional communities, reporting distortion usually runs downward. A chess career moves through four gatekeepers: coach, arbiter, organiser, investor. The direct cost of a complaint is the next invitation, the next camp, the next team list.

That makes 75 percent a floor rather than a ceiling. The worst experiences sit outside the percentage, in silence. This is an old lesson of mine. I tested hot takes on a chess stream in 2026 — the numbers never lied again after that. That broadcast carried 23 minutes of dead air and nobody had counted it. What is not counted does not get fixed.

What should have been measured and wasn't

The real damage here is not ACPL or engine match rate. It is the route through which bias and harassment hit performance. The route is direct: event selection.

A player who avoids mixed open sections is giving up the fastest rating accrual available. A player who stops travelling to a particular city or a particular organiser's event is watching her calendar shrink. A player who steps away mid-season has her preparation cycle broken. All of it shows up in rating. Rating never explains why.

All three are cheap to measure. Withdrawal rates by gender. Participation continuity across a cycle. Drop-out rates by age band between 15 and 20. Comparative error rates under time pressure. None of those appeared in the coverage.

Until those four numbers are public, 75 percent is a sentiment metric, not a performance metric. Sentiment sets ethics. It does not set budgets.

The jurisdiction question nobody touches

Much of today's harassment in chess happens where FIDE's authority is on paper and not in practice: platform chat, direct messages, stream comments, social replies, group chats.

This brings back 2026. Server failures in the FIDE Online Olympiad final forced India and Russia to be declared joint winners. Most coverage filed it as a feel-good COVID story. I pulled the match log, counted the disconnects, and wrote that India's first global team title came with an asterisk made of Wi-Fi. It was uncomfortable. Federation officials replied publicly for the first time.

Same logic here: if this survey covers online behaviour — and it very likely does — then FIDE is publishing data on conduct it can barely enforce against. Rules will be written that platforms will not apply. That gap belongs to governance, not to the survey, and it barely surfaced in the reporting.

The transfer-window question: who pays the bill

Money in chess is moving faster than ever this cycle. The Global Chess League's Mumbai debut in December 2026, franchise contracts, appearance fees, broadcaster sponsorship. Capital is flowing toward franchise branding faster than it reaches players — a separate argument, connected to this one.

If FIDE builds a policy package on this number — an independent complaints body, arbiter training, conduct compliance as a condition of organiser licensing, a legal support fund for women players — which account pays? The franchise owners? The national federations? Or the player who filed the complaint?

This is where the statistic and the contract face each other. A safety policy carries a cost line, and the cost line is how you tell whether an institution is serious or merely visible.

Where I could be wrong

The strongest pushback is that the stratification I demand dies with the data. The population of women chess players is small and split across national federations. Publish country-level breakdowns and a single country's respondent pool falls to ten or twelve. Their experiences can be read individually, but they cannot be selected — and in this profession, being identified is the single largest risk.

That objection is legitimate and I accept it fully. An unstratified number may have been a methodological necessity, not laziness. But necessity gets stated. One line could have read: to protect confidentiality we are not publishing country-level data. That line was absent. That is why the question stays open.

A second objection lands on me: over-applying the Kolkata lesson. The 2026 edition became my template for venue calibration, but Chennai's Anand-era nostalgia and Mumbai's 2026 debut work differently. Not every city hands you crowd tape. The same applies to this survey — treating one national reading as the global benchmark would be the same error as declaring a sport dead because one hall looked empty.

A third is sharper: do I actually want the sample size, or am I looking for an exit from the number? Every change of the past decade — in sport, in newsrooms, in corporate reporting — usually arrived before methodological completeness, driven by pressure. If the 75 percent figure produces an independent complaints body tomorrow, whether the sample was 300 or 400 will not matter. I concede that. My objection is not to the number. It is to using it as noise.

I have watched three eras of sports media, and the whisperers lost every single time. Chess broadcasting has carried one disease for four decades: it believes the audience deserves a hush. The same thing is happening to this survey. The number was said loudly, then immediately converted into a whisper. Stop whispering.

Takeaway: one date, one bet

My prediction is testable. If FIDE does not publish three things together within the next two publication cycles — the full technical document (sample, period, geography, method), gender-disaggregated withdrawal and drop-out data, and a complaints mechanism with a costed budget — then 75 percent becomes a slide. A number shown at every conference and forgotten at every budget meeting.

There is more money in chess than before. The question is not what three in four women endured; that has been stated. The question is how many of the four organisers, coaches and franchise owners will pay their share of the bill. Until that answer arrives, remember: Kolkata didn't shake hands. It watched. We were the ones who stopped counting.