The Empty Square on the Scoreboard: Chess Analysis When the Evidence Disappears

The white strip on the third monitor

In November 2026, in Singapore, game 14 of the world chess championship match between Gukesh D and Ding Liren moved into the endgame. I was sitting in Shenzhen with three monitors in front of me: a live stream window, an engine analysis panel, and a real-time Elo board. Around move 40, the third monitor froze. A blank strip, no new data point, no rating change, not a single line of annotation.

Seven minutes later I realised I was reciting metrics that did not exist. Ding Liren's endgame accuracy — I "estimated" it at about 92 percent. Gukesh D's win rate with the white pieces — I "roughly remembered" it as 58 percent. No table confirmed either figure. They came from memory, from feeling, from a wish that the result would follow the direction I had imagined before the game began.

When the data stops speaking, the analyst is the first person to start making things up. A blank page frightens no one. It only makes people want to write something convincing on it.

This is worth retelling, because in twenty-eight years of watching this industry I have never seen a more dangerous analytical error than filling a blank with content that sounds plausible. That error makes no noise. It does not crash a ranking, it does not cancel a tournament. It only makes a generation of readers believe things that were never verified.

My career began with a blank square

In 2026 I started out as a chess player and tournament organiser, then moved into chess media. Seven years of commentary for VTC taught me something no school teaches: audiences do not remember what you got right. They remember what you said with certainty.

In 2026, at the age of 35, I was a senior specialist at a sports data company in Shenzhen. My first assignment was to analyse the performance of the Brazilian striker Luis Fabiano while he played for Tianjin Quanjian. I reconstructed all 22 of his Chinese Super League goals and cross-checked them against expected goals (xG) and touches inside the penalty area.

The result silenced the meeting room. Fabiano's actual efficiency ran 18 percent below expectation. He scored heavily, but he scored in a predictable way: an outsized dependence on set pieces. I presented the data to the club's leadership and argued that their attacking system had become easy for any opponent to read. The club changed its tactics and signed a younger striker with better pressing numbers.

A Chinese club taught me that data is not the destination, it is the walking stick. That lesson arrived early and at the right moment, before I could start believing that spreadsheets could replace judgement.

From then on I moved entirely into deep analysis of xG and PPDA rather than conventional football reporting. And I formed one professional habit: every claim must carry a verification step.

In the scouting trade I now work in, that step has a name. Before you value a footballer — or a chess player — you must decompose raw material into citable events. What is the article's headline. Which source published it. On what date. Is at least one person or organisation named. Is at least one verifiable factual claim present.

Without those four things, every layer of analysis above them is decoration. You can build an eight-part report, full of charts, full of terminology, full of risk sections — and the whole report is still empty, because the base layer returned nothing.

That is the kind of error I mean. It does not live in the conclusion. It lives in the fact that people write conclusions anyway, without any data to start from.

The Niemann – Carlsen affair: a blank filled with a verdict

On 4 September 2026, at the Sinquefield Cup in St. Louis, Magnus Carlsen lost to Hans Niemann with the white pieces. After that game, Carlsen withdrew from the tournament. He posted a short line on social media implying that if he spoke, he would be in serious trouble.

In October 2026, the online chess platform Chess.com published a long report raising concerns that Niemann had likely cheated in more than 100 online games. A lawsuit was later filed by Niemann in the state of Missouri against Carlsen, Chess.com and several related parties.

On the data side, three points need separating. Evidence of cheating in over-the-board games was never published publicly. Every conclusion about Niemann at that time rested on inference from an online sample, which operates under completely different conditions. And the gap between those two kinds of data is precisely where public opinion built its verdict.

I wrote a piece in September 2026. I used the phrase "almost certainly". Three months later I had to retract that phrasing. When the data does not lie, we are the ones lying to ourselves. The evidence never appeared, but the story was already complete: a young player, a wounded champion, a platform in the middle. That narrative structure was so appealing that nobody wanted to wait for data.

My job is valuation, not jury duty. But I let myself be pulled into the second role, simply because a blank needed filling.

Elo thresholds and stories without data

In chess, certain thresholds become icons: 2700, 2750, 2800. A player crossing 2800 in classical Elo is treated as a historic event. Live ratings update after every game, and every brush with a threshold means headlines get rewritten.

The problem is sample size. A live rating is a violently oscillating line. A player can spike to 2795 after three straight wins, then return to 2770 after a seven-round event. If you read that oscillation without checking the sample size, you will see "turning points" appearing and vanishing every month.

Three metrics I always demand before saying anything about a player:

• ACPL — average centipawn loss per move against the engine's best choice. Lower is better.
• Engine match rate — the share of moves matching the engine's top pick.
• Opening novelty count — moves never previously recorded in the game databases.

It sounds strict, but here is the part that matters: a good ACPL across nine games says nothing about long-term class. It says the player had a good tournament. Those are two different sentences, and a great many analyses merge them into one.

I also always build the head-to-head table. Not to crown a winner, but to find the "bogey opponent" — a match-up skewed far beyond what the Elo gap predicts. There are pairings where a 5-1 record persists for years while the rating gap sits under 30 points. That data cannot explain the cause, but it points to where the questions should go.

Format produces the data, not the other way round

The tournament system is where data is manufactured, rather than where data is reflected.

The road to a world championship match passes through four doors: the FIDE World Cup, the Grand Swiss, the Grand Chess Tour, and the rating spot. The wild card is the fifth door, and the most contested, because it is decided by people rather than by a table.

The 2026 season gives a tidy example. At the World Cup in Baku, Magnus Carlsen beat Praggnanandhaa R in a tiebreak after the classical games were drawn. That same year, Vidit Gujrathi won the Grand Swiss. In April 2026, in Toronto, Gukesh D won the Candidates with 9 points from 14 rounds and earned the right to play for the title.

Three routes, three different kinds of data. The World Cup is a knockout, where one bad game can erase a year of preparation. The Grand Swiss is an 11-round Swiss, where squad depth matters more than peak form. The Candidates is a round-robin, where every opponent knows every other to the last opening variation.

And then there is Armageddon — the format I consider the clearest lesson in how format generates data. White gets more time, Black wins on a draw. Mathematically it is a nearly balanced structure. Psychologically it is an entirely different game.

The draw rate in elite classical chess typically hovers around 60 percent. That figure gets read as proof that chess is boring. It can also be read the opposite way: when two elite players have equal preparation quality, a high draw rate is the inevitable result of quality, not of dullness. You cannot demand maximum accuracy and then be disappointed that both sides achieved it.

The prize fund for the 2026 world championship match in Singapore was 2.5 million US dollars, split according to win-loss proportions. This is important financial data, because it determines how professional the top layer of players can be. A season in which the reward for the world title is smaller than the reward for a Candidates place can invert the entire competitive incentive structure.

The pipeline from training academies to platforms

I map the chess industry's transmission chain into three tiers. Upstream is the training pipeline and the talent supply. Midstream is the events, the players and the platforms. Downstream is content, commerce and derivative markets.

Upstream, India is the phenomenon nobody can ignore. In September 2026, at the Olympiad in Budapest, India took gold in both the open and the women's sections. That is the result of a decade of investment in academies, coaches and junior circuits, rather than of one exceptional individual.

Uzbekistan is a different model. At the 2026 Olympiad in Chennai, Uzbekistan won gold in the open section. A year earlier, in Warsaw in 2026, Nodirbek Abdusattorov won the World Rapid Championship at the age of 17. No vast financial base, no dense academy network. A generation of talent arrived at once, and resources were concentrated on exactly that group.

Midstream, online platforms have become genuine competition infrastructure. Chess.com's Titled Tuesday runs weekly, gathers titled players, and generates more game data than any traditional tournament. Alongside it come enforcement waves, when large numbers of accounts are closed after automated reviews.

Platforms have become regulators in practice, while formal authority still sits with FIDE. The distance between those two entities is one of the biggest risk points in chess today, and it appears in almost no financial report.

Downstream, federation transfer is a real market. After 2026, many Russian players moved to compete under other flags, dragging with them changes in qualification places, sponsorship rights and media positioning. It is the kind of move anyone in the football transfer world recognises instantly: changing a single line of administrative data can change the market value of a human being.

When a small country develops a strong generation, the reward is rarely a stronger domestic tournament circuit. It is that larger structures come and take them away. A training pipeline's success becomes the opening act of a different dismantling. I have seen this exact mechanism at small football clubs, and it works no differently in chess.

Transfer-window noise and buried signal

We are in the middle of a transfer window, and this is the worst noise-to-signal ratio of the year. I rank rumours by four levels of evidence. Level one is an official announcement, tied to a medical and a signed contract. Level two is two independent journalists with a track record of accurate reporting. Level three is the voice of an agent, who always has a stake. Level four is aggregator sites copying rumours from one another.

Most readers consume at level four and do not know it. In chess, the same scale applies almost intact. Rumours about a player switching federation, about a tournament landing the world number one, or about a club in a team league signing someone — all of them can be sorted into those four levels.

Contract structure, prize funds and waiting periods are the real story. Readers need a reliability filter, not another news item. And as clubs begin to list and must publish financial statements, the pressure of the reporting period will weigh on sporting decisions more than any coach. Fan emotion becomes a line on the balance sheet, and that line has a maturity date.

A professional glossary cannot rescue an empty table

I keep a daily glossary: Elo, live rating, performance rating, ACPL, engine match rate, opening novelty, opening preparation, tiebreak, Armageddon, time trouble, draw rate, the Candidates, the Grand Swiss, the World Cup, the Grand Chess Tour, round-robin, Swiss system, rating spot, FIDE, federation transfer, over-the-board play, Titled Tuesday, seconds.

The glossary is necessary. Without it you cannot read a technical report. But it does not generate data by itself.

This is what I learned the most expensive way. It took me three months to learn that a beautiful chart is worth less than a correct process. In 2026, at the World Cup finals in Russia, I predicted Germany would defend the title. My basis was possession share and pass completion in qualifying. Germany went out in the group stage after a 0-2 loss to South Korea.

My data was not arithmetically wrong. It was wrong in its selection. I measured what was easy to measure instead of what was decisive: pressure conversion, the speed of wide attacks, the quality of second balls. Three weeks later I rewatched all 48 group-stage matches, teaching myself metrics I had never used before.

My article structure changed from then on. I always reserve a section to question my own initial assumption. If that assumption does not survive, the rest of the piece has to be rewritten.

An empty data table cannot be rescued by vocabulary. It can only be handled correctly: state clearly that it is empty, stop, and go back to the base layer to find where the data went missing.

When contrarianism becomes a personality

I have a natural instinct to hunt for the hole in an overly polished chart. That instinct is useful in this trade. It is also a trap.

The professional sceptic and the instinctive sceptic differ on one point: the first attacks the process, the second attacks the conclusion. The first accepts that he might be wrong. The second only needs to look sharp.

In chess, the dangerous variant of this trap is the argument that every result is suspect. A player who performs too well is suspected. A player who performs too badly is suspected differently. When suspicion becomes the default starting point, it stops being analytical method and becomes a moral position disguised as technique.

I hit that limit in the 2026 affair. I was right to say that anti-cheating systems in over-the-board events have many holes. I was wrong to let that correctness turn into a verdict about an individual.

There is another parallel worth thinking about, this time from football. The millimetre offside line has changed the attacking instinct of entire teams. Strikers began slowing down, waiting, calculating. The referee shifted from running the match to editing it, with the authority to erase a goal using a computer frame.

In chess, that editing tool has existed longer and is stronger: the engine. A move can be labelled a mistake the instant it is played, by a machine that knows nothing about time pressure, about head-to-head psychology, about a human having to choose between two equally correct options. But without the engine, we would have nothing to cross-check against.

This is the central paradox of my trade. The stronger the tool, the easier it becomes to believe the tool has replaced judgement. A Chinese club taught me the opposite years ago, and I still have to remind myself of it every time a table looks too perfect.

An early-warning system

In 2026, when major competitions were suspended and stadiums stood empty, most sports media people lost income at the same moment. I stayed home and started a different project: a ten-year analysis of Premier League transfer data. The most notable result was a correlation around the adaptation of Brazilian wingers — those who had played in Portugal before moving to England showed a success rate roughly 42 percent higher.

I stated clearly in the piece that this was correlation, not causation. It could be a language effect. It could be an effect of adapting to European playing styles. It could simply be a statistical illusion from a sample large enough to look meaningful but not large enough to conclude anything.

Correlation is not causation, and that is why I write the contrarian section before the conclusion. The piece was widely shared, and it carried me through a period of lost income during the pandemic. But its real value lay elsewhere: it forced me to build a long-horizon tracking system instead of issuing a short-term prediction.

After 2026, I stopped believing in predictions. I believe only in early-warning systems.

An early-warning system works on a completely different principle from a prediction. It does not say who will win. It says a metric has just left its normal range and needs watching. It accepts that most signals will lead nowhere. It does not need to be right often; it only needs not to miss real change.

That is also how I read chess today. The era of a single absolutely dominant champion has given way to a group of players separated by ever-narrowing Elo gaps. Gukesh D became world champion at 18 in December 2026, after Ding Liren had held the title for less than two years. Praggnanandhaa R reached the World Cup final in 2026 at 18. Alireza Firouzja is the youngest player ever to cross 2800 Elo.

That data does not tell me who the next champion will be. It tells me the dominance cycle has shortened, that the training systems in India and Uzbekistan are producing talent faster than tournaments can absorb it, and that the cost of a single mistake in a single game keeps rising.

And it tells me one more thing, which I have to write down because it is the reason I sat in front of three monitors on a November night in 2026: most media crises in sport do not begin with a wrong data point. They begin with a blank that somebody filled too quickly. This transfer window will produce thousands of such blanks. The question for readers is not who is going where, but whom you are reading — and whether that person is willing to leave a square empty when there is nothing yet to write.