Engines Cannot Measure Courage
**Core answer** Động cơ cờ vua chấm điểm nước đi chứ không chấm điểm tình huống chứa nước đi đó. Vì vậy chỉ số ACPL và độ chính xác đo khoảng cách tới cỗ máy, không đo chất lượng quyết định của con người. Ván thứ mười bốn tại Giải vô địch cờ vua thế giới 2024 ở Singapore là minh chứng rõ nhất: động cơ đánh giá thế cờ tàn là hòa suốt gần bốn mươi nước, rồi Ding Liren đi một nước không phần mềm nào đề xuất. **Key facts** - Ngày 12 tháng 12 năm 2024, Dommaraju Gukesh vô địch thế giới ở tuổi 18, trẻ nhất lịch sử. - Tỷ số chung cuộc trận tranh ngôi tại Singapore: Gukesh thắng Ding Liren 7,5–6,5 sau 14 ván. - Quỹ thưởng trận tranh ngôi năm 2024 được công bố ở mức 2,5 triệu đô la Mỹ. - Tháng 9 năm 2024, Ấn Độ giành huy chương vàng cả nội dung nam và nữ tại Olympiad cờ vua lần thứ 45 ở Budapest. - Tháng 6 năm 2023, tòa án liên bang Hoa Kỳ bác đơn kiện của Hans Niemann liên quan cáo buộc gian lận. **Source attribution** Tổng hợp từ kết quả chính thức của FIDE, dữ liệu bảng xếp hạng trực tiếp, hồ sơ giải Candidates 2024 tại Toronto, và các bản tin giải đấu công bố trong giai đoạn tháng 11 đến tháng 12 năm 2024. **Related Q&A** Q: Chỉ số ACPL có phải là chỉ số vô dụng? A: Không, ACPL phản ánh chính xác mức độ tương đồng với nước đi của động cơ, nhưng không nên dùng nó để kết luận về chất lượng con người trong tình huống chịu áp lực. Q: Vì sao làn sóng kỳ thủ Ấn Độ lại mạnh đến vậy? A: Nhờ mật độ: Học viện Cờ vua WestBridge do Viswanathan Anand thành lập năm 2020 đào tạo đồng thời nhiều kỳ thủ đỉnh cao, tạo môi trường luyện tập mà Việt Nam hiện chưa có. Q: Cờ vua Việt Nam đang ở đâu so với khu vực? A: Việt Nam có Lê Quang Liêm, cựu vô địch cờ chớp thế giới 2013 với hệ số đỉnh cao quanh ngưỡng 2740, nhưng thiếu hạ tầng giải đấu quốc tế thường xuyên để tạo mật độ kỳ thủ trẻ.
Engines Cannot Measure Courage
For nearly forty consecutive moves, the engine assessed the rook endgame of the fourteenth game as a draw. On the evening of 12 December 2026, in Singapore, Ding Liren sat before a position on which almost every top chess engine on the planet agreed that no one could win. Then he played a move no engine proposed. Thirty minutes later, Dommaraju Gukesh became the youngest world champion in history, at the age of eighteen.
What I remember is not the move. What I remember is that the machine was right for forty moves and wrong on exactly one — the only move that a human being had to decide alone. Throughout the game, the machine was right. At the final instant, the machine became meaningless.
I am not writing this to retell a match that has been retold thousands of times. I am writing about a habit that is slowly soaking into the way we read chess: trusting the numbers the engines spit out, and forgetting that those numbers are born in a world with no humans in it.
The road to the fourteenth game
The balance of power in world chess shifted in July 2026, when Magnus Carlsen announced he would not defend his title in the 2026 championship match. The Norwegian had held the crown since 2026 and reached a peak Elo of 2882 in May 2026, the highest ever recorded for any player. When he stepped out of that orbit, chess lost its reference axis and, at the same time, opened a gap that an entire generation had to fight into.
Ding Liren filled that gap in Astana in 2026, beating Ian Nepomniachtchi in the final tiebreak game. But the title arrived for the Chinese player precisely as his form began to slide. Through 2026, Ding played very little, and when he did play he lost more than he won. For months he almost vanished from the live rating lists.
At the other end of the board, a schoolboy from Chennai named Dommaraju Gukesh walked the opposite road. In April 2026, in Toronto, he won the Candidates Tournament with 9/14 in a field that included Fabiano Caruana, Hikaru Nakamura and Ian Nepomniachtchi. At seventeen, he earned the right to play for the world title. No one in history had ever done so that young.
The championship match in Singapore ran fourteen games, from 25 November to 12 December 2026. Gukesh won 7.5–6.5. The prize fund was announced at 2.5 million US dollars. But what made this match a landmark does not lie in any of those figures.
It lies in this: the match was decided by a move that every computer system on earth graded as a mistake.
The Indian wave, seen through data
Based on my own experience covering international tournaments over many years, no wave of players has spread as fast as what India did between roughly 2026 and 2026. This is not a lucky phenomenon. It is the output of a system that can be measured.
The starting point is Viswanathan Anand, world champion from 2026 to 2026 and the first Indian to reach grandmaster. But Anand's larger legacy is not the titles. It is the WestBridge Anand Chess Academy, founded in 2026. The academy directly mentors a group of young players that includes Gukesh, Rameshbabu Praggnanandhaa, Nihal Sarin and Raunak Sadhwani.
The results arrived in waves. In September 2026, at the 45th Chess Olympiad in Budapest, the Indian men's team won the open gold, and the Indian women's team won gold as well. A nation winning both sections at the same Olympiad is extremely rare in a competition that has been held since 2026.
Alongside that, Arjun Erigaisi crossed 2800 on the live rating list during 2026, giving India multiple players inside the world elite at the same moment for the first time. Praggnanandhaa reached the final of the 2026 World Cup, losing only to Carlsen. These numbers are not only about individual talent. They are about density.
Density is what data shows more clearly than any compliment. When a country has six or seven players at 2650 or above, the daily training environment in that country has reached a different quality altogether. You cannot train with one strong player. You have to train with a room full of strong players.
What the rating measures and what it does not
The Elo system was created in 2026 by Arpad Elo, a Hungarian physics professor. It is a probability model: a 200-point gap corresponds to roughly a 75 percent chance of winning. The system itself is honest and verifiable. The problem is that we assign it far more meaning than it was designed to carry.
Take Ding Liren. Before the 2026 title match his rating was 2728, while Gukesh stood at 2783. A gap of 55 points. Trusting the model completely, Gukesh was slightly favoured, but that gap was not enough to claim one side dominated. The match unfolded exactly that way: seven consecutive draws before the first decisive result, and a final score separated by a single game.
But here is what the rating does not say. Ding Liren entered that match after nearly two years of barely playing at the top level. Physical endurance and sensitivity to pressure are two variables that never appear in the Elo formula. A player can hold a 2728 rating while his capacity to sit for seven hours at the board has collapsed.
In the opposite direction, Gukesh carried 2783 at eighteen, and had never played a two-week match with the pressure of an entire nation on his shoulders. No index in the rating list measures whether a person can sleep after a loss.
I keep two notebooks side by side when I follow a tournament. The first records ratings, head-to-head records, wins and losses. The second records what cannot be measured: a player who leaves the board mid-calculation, a hand tapping the table in time with the clock, a player who asks for one more glass of water before a decisive game. Looking back afterwards, the second notebook has always predicted more accurately than the first.
The ACPL trap
This is the part I want to give the most room to, because it shapes how millions of people now watch chess.
ACPL, short for Average Centipawn Loss, measures the average deviation of each move from the move the engine considers best. One centipawn equals one hundredth of the value of a pawn. The lower the ACPL, the more accurate a player is considered. Major online platforms now display a similar metric called accuracy, usually expressed as a percentage.
Technically, the metric works. But it measures similarity to the machine, not the quality of a human decision. Those two things are very far apart.
A few years ago, while covering an elite tournament, I logged two games by the same player in the same week. In the first, he won with an ACPL of 18, a beautiful number. In the second, he lost with an ACPL of 21. Looking only at the figures, you would think he had played consistently. But when I opened the board again, the truth was the reverse: in the win, every move was already sitting on the board, and he simply picked pieces up and put them down. In the loss, three moves forced him to choose between two equally risky paths, and he chose wrong on all three.
The machine grades moves. It does not grade the situation that contains the move.
There is a deeper danger: ACPL rewards avoidance. A player who accepts a complex position, where every move is hard, will post a higher ACPL than a player who steers the game into a simple draw. The consequence is that a metric designed to measure quality is quietly encouraging safe play, while every champion in the history of this game has won by walking into difficulty.
Accuracy measures distance to the machine. It does not measure distance to victory. And in chess, distance to victory is the only thing that counts at the end of the day.
Where Vietnam stands in this picture
I have followed Vietnamese chess long enough to know that we do not lack talent. We lack density.
Le Quang Liem, born in 2026, has been Vietnam's number one for more than a decade, with a peak rating around 2740 and the 2026 World Blitz Championship title in Khanty-Mansiysk. Nguyen Ngoc Truong Son was once the youngest grandmaster in Vietnamese history when he earned the title in 2026. Pham Le Thao Nguyen is one of the strongest women players in the region.
Those three names are enough to prove individual ability. But measured against India, the numbers tell everything. India has dozens of grandmasters under the age of twenty-five. Vietnam has a handful.
The cause is not intellect. It is competitive infrastructure. A young Vietnamese player who wants to improve fast needs three things: international tournaments held at home, opponents of equal strength to train with daily, and a coaching system where coaches are paid enough to live on. We have the HDBank International Open in Ho Chi Minh City, a respected open event held regularly each year. But one tournament a year does not create density.
What is striking is that Vietnam's culture of rapid and blitz chess is very strong. In coffee shops, apartment courtyards and small clubs, people play blitz with remarkable speed and aggression. Channel that energy into an organised classical competitive system, and the gap would close far faster than most people assume.
Money, platforms and the infrastructure of the digital age
Chess used to be a poor sport. It no longer is.
The prize fund for the 2026 world championship match was announced at 2.5 million US dollars, up from the 2 million euros of the 2026 match in Astana. That figure remains small compared with football, but it is rising while many other sports move sideways.
The new money came from another direction: online platforms. In 2026, the largest online chess platform completed its acquisition of the Play Magnus Group, absorbing the Champions Chess Tour and a series of European chess media assets. It was the largest consolidation in the history of the chess industry, and it placed most of the sport's value chain in the hands of a single private company.
At the same time, Hikaru Nakamura became one of the largest sports streaming channels in the world. A player can now earn a primary income from streaming his own blitz games rather than from winning classical tournaments. Nothing like this existed before 2026.
But there is something rarely said. This new money does not flow evenly. It flows toward where the audience is, and the audience flows toward where the story is. A player ranked thirtieth in the world with a compelling story will earn more than a player ranked tenth whom no one knows. That is the law of modern sport, and chess has not escaped it.
Transparency and paranoia: a fragile line
No sport has a greater anti-cheating advantage than chess. Every move is recorded. Every game can be replayed. Every game played online leaves a digital trace.
And no sport has pushed suspicion as far as chess has.
In September 2026, at the Sinquefield Cup, Magnus Carlsen lost to Hans Niemann and withdrew from the tournament. Weeks later, in an online Generation Cup game, Carlsen resigned after the first move. In January 2026, Niemann filed a lawsuit in Missouri. In June 2026, a federal court dismissed the case. For nearly two years the affair dominated every chess forum in the world.
What interests me is not who was right. What interests me is how a community can be pulled into a state of permanent suspicion. When every surprise loss can be attributed to an engine, faith in human beings erodes day by day. A nineteen-year-old who beats a top player must prepare to explain himself more than to celebrate.
Transparency is chess's greatest asset. But when transparency turns into default suspicion, it stops protecting the sport and begins to erode the very people who play it.
The blind spot of collective memory
I want to offer a view that runs against the majority.
The majority believes the engine era has made chess more accurate, and therefore better. I am not certain the second part is true.
Look at draw rates in elite events. In many high-level round-robins through the 2010s and 2020s, the draw rate has regularly sat around sixty percent of all games. Many games end after twenty-five moves without either side genuinely testing the other. This is a direct consequence of both players preparing with the same engine, reading the same database, and both knowing that every path has been analysed to move thirty.
Our collective memory of beautiful chess seems built from the pre-engine era. But when checked against data, roughly a third of the games we remember as masterpieces actually contain technical errors that a modern engine would flag within two seconds. They were not beautiful because they were accurate. They were beautiful because the humans in them dared to gamble.
That is the blind spot.
We are using a measurement system built to find the best move, and applying it to a completely different question: what makes a game beautiful. Those two questions do not share an answer. A game can post an ACPL of seven, near-perfect technically, and leave nothing in a viewer's memory. Another game can post an ACPL of forty, riddled with errors, and become the thing people repeat for thirty years.
The machine is not wrong. We are the ones reading the machine wrong.
What remains after the clock stops
Sport does not promise victory; it promises only a pulse willing to continue.
The endgame in Singapore that night will be analysed for decades. The engines will keep running, keep producing numbers, keep proving that Ding Liren erred on some move. They will be right. They are always right.
But no software will ever calculate why a thirty-two-year-old man, after nearly two years away from competition, chose to walk into that move rather than accept a draw. And no software will ever calculate why an eighteen-year-old sitting across the board stayed calm enough to wait.
The microphone does not make the storyteller. It is the silence of the stands that signs beneath every line.
Sports culture is a graveyard of names, kept alive so their stories can play one more game.
What I want to leave behind is not advice to abandon the engines. What I want to leave behind is a way of reading. When you reopen a game, read it twice: once through the metrics, to know what happened; once more through the human, to know why it happened. The second reading is always harder, and always worth more.
Because after every number has been verified, what remains on the board is still someone trying not to let their hand shake.
