Fritz in Hanoi: What Vietnamese Chess Learned from the 2026 Philadelphia Shock

On the night of February 10, 2026, in Philadelphia, Garry Kasparov sat motionless for a long time before signing the scoresheet for game one. IBM's Deep Blue had just beaten the world champion. In Hanoi, in a small chess club on Hang Bai Street, a 14-inch television replayed a clip less than a minute long: no board visible, only a man with his face lowered before a computer cabinet that came up to his chest. Deep Blue read up to 200 million positions per second. Kasparov, at the peak of his career, could see a few dozen candidate lines per minute. Six games from February 10 to February 17, ending 4-2 in favour of the human. That score was not what silenced the room on Hang Bai Street. The first game was. I was twenty-seven that year, nine years into commentating on chess for Vietnam Television. I once believed in feeling, until a number knocked on my door at three in the morning. The number was 200 million. It forced me to rewrite the way I look at a chess game. Vietnamese chess was at a different turning point then. Dao Thien Hai had just earned the international grandmaster title, the first Vietnamese to reach that level. The national team was preparing for the Olympiad in Yerevan. But training conditions were still those of the previous decade: a few handwritten books, one coach, a handful of sparring partners, and a chess clock. A young player in Hanoi in 2026 got through a few hundred serious games a year. A player in Moscow got through three times that. The gap between Hanoi and the rest of the chess world was not talent. It was the number of games played, and the number of games written down. That same year, a German software company put Fritz on the market, a chess program that ran on a personal computer. Fritz was sold as a training partner, not an opponent. A user could set the strength level by grade, ask the machine for a hint, force it to play one chosen opening, and save every game for review. A year earlier, in Hong Kong, Fritz had won the World Computer Chess Championship. Kasparov's defeat in Philadelphia was the hot story. Fritz was the long story. Most articles this year will tell it one way: machines are about to overtake humans, humans had better prepare. I read the scoresheets of the six Philadelphia games three times and saw something else. A computer beating the world champion does not change how people learn chess. A computer becoming the person you train with does. And what Fritz changes is not the quality of a player's moves. It changes how many times a player is allowed to be wrong. Before this, a student in Hanoi who lost a game went home and tried to remember it, got as far as move twenty, and then guessed. One lost game produced one lesson, and that lesson was often wrong because the memory was wrong. With the software, the game is recorded move by move, and the mistake is flagged within hundredths of a second. One lost game produces twelve lessons, and all twelve are right. This is where I want to slow down. The machine finds a mistake thousands of times faster than a human. It does not teach a human to find the mistake. Those are two different things, and the distance between them is where a player is made. There is something else only software can do. Until now, a strong player in a provincial town could be better than a player in the capital and nobody would know, because nobody recorded his games. The machine records everything. Some players are forgotten, but the data never forgets them. Looking at how Fritz is built, the most valuable thing for a Vietnamese player is the adjustable strength. Playing an opponent two hundred Elo above you teaches you something; playing one eight hundred Elo above you only teaches you endurance. Just as valuable is repetition: ten games in the same endgame teach more than ten games in ten different openings. And behind all of it sits a diary that never forgets. The easiest thing to miss this year is a correlation read as causation. People will say: as machines get stronger, players get stronger. The number of games played against machines rises, and a new generation of players will appear. Games do not produce strong players. A man can play five thousand games against Fritz in one year and still lose to a forty-year-old club player in a queenless endgame. The reason sits in the fact that the machine has no model of its opponent. It evaluates the position. It does not evaluate the person sitting opposite. A position the machine scores at 0.00 for thirty moves can still be won, if the player on the other side has been sitting for six hours, has finished his water, and missed a chance at move twenty-five. No number in that software measures any of that. There is another risk that few people mention. When every player trains with one piece of software, every player walks into one opening. In 2026, in Vietnamese tournaments, people played the Sicilian Defence because a book said so. Ten years from now, they will play one Sicilian line because a machine scores it 0.15 instead of 0.10. The variety of human chess will narrow, and that narrowing will be called progress. A move is not noise. It is a question that the numbers are whispering. Leaving that room on Hang Bai Street that night, I was not thinking about Deep Blue. I was thinking about the clubs in Nam Dinh, in Hue, in Can Tho, where a fifteen-year-old might have a training partner who never tires, never loses patience, and never charges a fee. Within a year or two, a machine will win an entire official match against the world champion. The question of 2026 is not when. The question is whether, before that happens, we use this tool to close the gap or to replace the teacher. I light a candle for the data. But I always let the flame of feeling light the way for the question.