Modern Chess and the Trap of Unverified Data

On the night of December 12, 2026, in Singapore, Game 14 of the World Chess Championship reached the endgame. Ding Liren, playing White, needed only a draw to force a tiebreak. But on move 55, the reigning champion made a losing rook move. Gukesh Dommaraju, just 18 years old, did not miss it. A few moves later Ding resigned, and Gukesh became the youngest world champion in history, breaking the record Garry Kasparov set in 2026 at the age of 22. I watched the game on a screen in Saigon. What made me pause was the engine's evaluation bar swinging from 0.0 to more than +2.0 in barely two seconds. A single number decided the fate of an entire two-year cycle. Within hours, hundreds of analyses flooded forums and social media. Most were unverified; many were simply copies of the engine's evaluation line with a few comments added to sound professional. I recount this because it lands exactly on what I have done for eight years — reading a game and trying not to fool myself — not to criticise the fans. When every move leaves a trace In two decades, chess has shifted from a game of intuition to a sport of data. An amateur today holds resources that world champions of the 1990s never dreamed of: databases of millions of games, Stockfish running on a phone, live FIDE rating lists, and platforms that analyse each move in seconds. That convenience creates a paradox. The more data there is, the blurrier the line between analysis and guesswork becomes. Fans are familiar with the numbers: classical Elo, rapid Elo, blitz Elo, performance rating, engine-match rate, and ACPL — average centipawn loss per move. Very few of them know where those numbers come from, or what happens when they are quoted without tracing the source. I once assumed chess was immune to this disease. Football has heat maps that can lie and pressing stats that can be inflated. Chess felt different: every move is recorded, results are clear, a win is a win and a loss is a loss, with no room for vague interpretation. But the deeper I dug, the more I saw that chess has the same grey zones. A game can be recorded move-perfect while the analysis written about it is completely wrong. In 2026, when the pandemic pushed chess online, I had to discard half of my old data. Post-lockdown chess is a different sport: faster, with a larger audience, and an explosion of online games so large that older statistical samples became meaningless. That was the first time I understood that more data is not automatically better. The push came from online platforms. When the 2026 season stalled, chess moved onto the internet and reached an entirely new audience, most of whom had never touched a board. But that same wave also brought shallow analysis, where a player is judged on three rapid games rather than a whole career. Eight layers of a decent analysis The framework I use for elite events has eight layers, and the first is always the technique of the game itself: opening system, move quality, engine-match rate. The most reliable indicator here is ACPL. A player who posts an ACPL under 20 in a classical game is considered near-perfect, and the figure can be rechecked with any engine. But ACPL only means something if we know which game it came from, in which format, and over how long. For the same player, ACPL in blitz is completely different from classical. Quoting a number without context is the fastest way to lie with statistics. The second layer is the player profile. Classical, rapid and blitz Elo tell three different stories. Someone can be world number five in classical and a blitz champion, or the reverse. Head-to-head records work the same way: seven wins in ten games sounds impressive until you learn all seven came in rapid while all three classical games were lost. Placing Gukesh in the wider picture, what stands out is not the title but the speed of his climb. He became a grandmaster at 12, the second-youngest in history to earn the title. His generation — Rameshbabu Praggnanandhaa, Arjun Erigaisi — forms the Indian wave reshaping elite chess. This is verifiable data, not a feeling. I once covered a youth event in Hanoi where an 11-year-old beat two opponents rated above 2200 in a row. The media immediately called him a prodigy. Six months later he had vanished from every tournament. Looking back at the data, those two wins came in blitz with three minutes per side — a sample far too small to conclude anything. The third layer is the tournament system. A world championship is nothing like an open Swiss, and both differ from an online event with prize money. To judge who is strong, you need the format, the number of rounds, the time control and the qualification route. The road to the title runs through the Candidates, the World Cup, the Grand Swiss, or a wild card — each route with its own criteria. The fourth layer is the competitive landscape, where I sort players into groups. At the top remains Magnus Carlsen's generation and the 2800 threshold very few reach. Below sits the challenger band around 2750. Then comes the rising class of young players. Placing someone in the right group says more than any praise. The fifth layer is rules and governance. Anti-cheating, format regulations, qualification conditions — this is the most sensitive part, because a false accusation can destroy a person's career. The sixth layer is risk. A young player can break out at one event and then disappear, because form in chess is more fragile than outsiders assume. Clock pressure, financial pressure, and the psychological weight of a world watching every move. More than a few talents burn out before reaching their peak. The final layer is how the chess industry transmits impact from the board outward: online platforms, the streaming wave, sponsorship deals, and public image. When a player becomes an internet phenomenon, money and attention pour in fast — and drain out just as fast. Federation transfer is chess, not a card game, yet many federations still prefer to flip cards. The trap called certainty This is where the story gets complicated. In September 2026, at the Sinquefield Cup in Saint Louis, Magnus Carlsen lost to Hans Niemann and then abruptly withdrew from the tournament. Days later, in an online game, Carlsen resigned after the first move. An implied cheating accusation swept through the chess world. By October, Niemann had filed a lawsuit seeking 100 million dollars. An investigative report later concluded that Niemann had likely cheated in online games but found no convincing evidence in over-the-board play. What matters is not the conclusion but how far public opinion ran ahead of the evidence. For weeks, the chess world split into two camps, each certain of its judgment, while the actual data sat with the investigators. I drew one principle from it: an evaluation bar can lie, but five consecutive moves drifting from the engine cannot. The evaluation bar is a composite number, distortable by position, by time, by how the engine is configured. But a repeated chain of errors in the same type of position is a signal that cannot be faked. Good data is measured by whether it can be refuted, not by how impressive it looks. No tactic is ever old; only the way we read a game expires. That line holds for football, and it holds even more for chess now that engines have changed how the whole world sees a game. What I keep Every time I sit down to write, I remind myself that the worst analysis is the one that states things which cannot be verified. A rating must trace back to the official FIDE list. A game must be cross-checkable against a database. A cheating accusation must have an investigative file behind it. Otherwise, what we are doing is just guesswork wearing a professional coat — and in a sport where every move leaves a trace, guesswork is the only thing that cannot be justified.