160 RPM Spin, 10 Milliseconds to React — Meet the Robot That Beat a World-Class Player

Sony's AI robot called "Ace" beat professional table tennis players in official matches. Not a simulation. Not a lab experiment. Actual games, actual umpires, actual rules — the same International Table Tennis Federation (ITTF) rules that govern Olympic competition.

What Happened, Exactly

Sony AI started testing Ace in April 2025. In those first rounds, Ace faced five elite amateur players — people who have trained for over 10 years and practice around 20 hours every week. Ace won three out of five matches. Across 13 games, it won seven.

Then came the professionals. Two Japanese league players, Minami Ando and Kakeru Sone, took the table. Ace lost both matches but managed to win one game. That alone was remarkable — no robot had ever done that before.

Sony went back to work. By December 2025, Ace had improved. It started winning against professionals too. In March 2026, Ace won matches against three professionals, including Miyuu Kihara — a player currently ranked in the world's top 25 in women's singles.

Think about that for a moment. A robot beat a top-25 player in the world.

Also Read: Amazon Put In $5 Billion. Google Put In $40 Billion. Both Are Betting on the Same Company

How Does Ace Actually Work

Ace doesn't win by brute force. It wins by reading the ball better than humans can.

The robot uses event-based vision sensors — hardware built by Sony itself, with a perception latency of just 10.2 milliseconds. That means it "sees" and reacts to the ball in about the time it takes you to blink. Except it never blinks.

The key advantage is spin detection. Table tennis is, at its heart, a game of spin control. Players put heavy topspin, backspin, or sidespin on the ball to make returns difficult. Ace returns successfully about 75% of spinning balls across a wide range of spin types. Human players struggle with consistency like that.

The robot learned using deep reinforcement learning — the same method that trained AI systems to master chess and Go. But this is different. This is the physical world, not a virtual one. The ball moves at speeds above 20 metres per second. Spin rates exceed 160 revolutions per second. And unlike chess, you can't "think" for a second — you have milliseconds.

Also Read: 4x Faster, More Accurate, Actually Useful — ChatGPT Image 2.0 is OUT!

Why This Matters Beyond the Ping-Pong Table

Peter Dürr, Director of Sony AI in Zürich and the project lead, put it plainly: "This research has shown that an autonomous robot can, in fact, win at a competitive sport, matching or exceeding the reaction time and decision making of humans in a physical space."

Sony AI's Chief Scientist Peter Stone went further: "This breakthrough is much bigger than table tennis. It represents a landmark moment in AI research, showing, for the first time, that an AI system can perceive, reason, and act effectively in complex, rapidly changing real-world environments that demand precision and speed."

That phrase — "complex, rapidly changing real-world environments" — is what engineers and policymakers should be paying attention to. The same capabilities that let Ace return a 160 rpm spinning ball in 10 milliseconds could theoretically be applied to surgical robotics, disaster response, warehouse logistics, or even battlefield applications.

Since the first "robot ping-pong" competition in 1983, researchers have chased this milestone. Decades of work, and now — finally — a robot that doesn't just participate but competes.

Ace probably won't join the professional circuit. That was never the point. Sony built Ace to push the edges of physical AI technology — and to learn what it would take to build robots that operate reliably in fast, unpredictable, human environments.

The lessons from a ping-pong table may end up mattering more in a hospital ward or on a factory floor than they ever will in a sports hall.

To read more such Bharat Explainer, click here.

About us:

News4Bharat is one of India’s fastest-growing digital news platforms, delivering timely updates and in-depth coverage across economy, policy, infrastructure, and national developments. Visit our homepage for more latest news and insights.

FAQs

Q1. Did Ace beat all professional players? No. In April 2025, Ace lost to both pro players. But by March 2026, after significant improvements, it was winning matches against professionals including world top-25 player Miyuu Kihara.

Q2. What technology does Ace use? Ace uses event-based Sony vision sensors, deep reinforcement learning, and a high-speed robotic platform with 10.2 ms perception latency.

Q3. Is Ace going to compete in official tournaments? No. The project was research-focused, not competition-focused. The goal was to push physical AI technology.

Q4. Was the match played under official rules? Yes. Matches were conducted under official ITTF rules with licensed umpires officiating.

Q5. What other applications could this technology have? Surgical robotics, disaster response, precision manufacturing, and human-robot interaction in safety-critical environments.