Sony AI's robot system 'Ace' defeats a professional in a table tennis match using official rules.
Sony Group's 'Ace' table tennis robot system has achieved a groundbreaking feat by securing its first victory against a top-class human player.
Ace Research Project | Sony AI
Sony AI announces groundbreaking research on real-world artificial intelligence and robotics - Sony AI
https://ai.sony/news/project-ace-press-release
AI robot defeats human in table tennis, playing against national tournament player, developed by Sony [The Age of AI]: Asahi Shimbun
https://www.asahi.com/articles/ASV4Q0RN8V4QUTFL01MM.html
Outplaying elite table tennis players with an autonomous robot | Nature
https://www.nature.com/articles/s41586-026-10338-5
Ping-pong robot Ace makes history by beating top-level human players | Reuters
https://www.reuters.com/sports/ping-pong-robot-ace-makes-history-by-beating-top-level-human-players-2026-04-22/
The sport of robot table tennis has existed since 1983, but developing the technology to catch a ball moving at a linear velocity of over 20 meters per second and a spin speed of over 160 rotations per second has proven extremely difficult.
Sony AI, Sony's AI research division, has developed a robotic system called 'Ace' that uses deep reinforcement learning through simulations that are faithful to the laws of physics, and can capture balls using a total of 12 sensors.
Sony AI collaborated with five elite table tennis players, each with over 10 years of experience, participation in national and regional tournaments, and who have trained an average of 20 hours per week, 5-6 times a week, over the past year, as well as two professional players active in Japan's professional league (T.League). Matches against the elite players were played as '3-game matches,' where the first player to win two games out of 11 points wins, while matches against the professional players were played as '5-game matches,' where the first player to win three games wins, and the matches were conducted according to the rules of the International Table Tennis Federation.
The results are as follows: Ace achieved 3 wins out of 5 matches against elite players and 0 wins out of 2 matches against professional players. Most previous competitive table tennis robots were designed with the premise of 'continuing rallies' in mind, and their performance in matches remained at an amateur level. Although Ace had a losing record, it demonstrated high performance by 'surpassing top players in interactive and real-time tasks.'
Furthermore, several additional matches were held after the paper was submitted. In a match held in December 2025, he defeated two elite players and one of two professional players. In March 2026, he played against three professional players and secured at least one win against each of them.
You can see footage of the actual matches in the video below.
Project Ace - YouTube
Professional player Mayuka Taira said, 'The strength of this robot is that it is extremely difficult to predict and does not show any emotions. Because you cannot read the opponent's reactions, you have no idea what kind of shot it dislikes or what kind of shot it struggles with, which makes playing against it even more difficult.'
Related Posts:
