The Robot Hangover

Submitted by stephen on Wed, 11/16/2011 - 09:54

When I first witnessed a table tennis robot in action I was excited. Who could ask for a better practice partner? The robot would send just the type of shots I needed to develop my game for match play. It was consistent and never got bored with repeating shots.

So I spent hours of diligent practice on the robot developing a great loop against underspin, and then I worked on offensive responses to the fastest topspin shots the machine could generate. Finally I stepped out onto the playing floor armed with confidence in the array of magnificent shots I’d perfected. I was ready to utilize my amazing new skills in match play. Surely this would be great!

But then the unthinkable happened. My opponent failed to cooperate: Balls began arriving in unpredictable sequences with variations in spin, speed and location I wasn’t able to read. I found myself flat-footed and unresponsive to my opponent’s behavior. In fact, I barely noticed my opponent at all. I was left with well-trained responses to shots that never came.

I tried to consciously separate robot practice from play to avoid this negative effect. But no matter how hard I tried, each time I would emerge from the comfort of robot training to find myself hopelessly inadequate during match play. The stupefying effect would last hours or days. I called it the “robot hangover.”

After many attempts I finally surrendered. I gave up using the robot. When there was no coach or practice partner available to deliver multi-ball I would simply practice service or not practice at all. Then recently Donn Olsen proposed an experiment with a novel PATT approach to robot training. He suggested that I envision an opponent striking the shot typeI had programmed the robot to deliver.

Imagine that the robot in front of you is not there, and in its place is a player hitting the ball to you?

This seemed a pretty far-fetched idea. Nevertheless, first thing the next day I set the robot to deliver topspin, then stepped back to receive. As the ball rocketed toward me I stroked and returned to ready position, trying to imagine a player counterattacking my shot with a fast loop. I reset myself and awaited the next ball, then struck again.

Quickly it became apparent that this just wasn’t right. No player I could imagine would counterattack so predictably, so I reset the robot to approximate the pacing of shots in a moderately high speed rally with some variation in placement. It worked! This was how match play occurred; realistic pacing of balls with spin type and placement reflected match play allowing me to envision my opponent, racket closed, counterattacking my loops.

It was mentally demanding to keep my “imaginary friend” active on the opposite side of the table, but I sustained for a half hour then went onto the court to play a human opponent. This was the real test. Repetitive drills, whether with a robot or a coach, have the tendency to build false confidence and poor judgment. How would I play?

Amazingly I experienced none of the ill effects I had experienced previously using the robot. While the process of imagining an opponent was not the same as reading the ball coming off an actual racket, I found that the mental energy used to imagine an opponent transferred effectively to watching a real opponent. The imagined surrogate had stood in very well for the judgment process. My match play was positively affected.

The next day I set the robot for underspin with the same pacing of shots. I tried, but could not imagine the underspining player on the far side of the net. What was wrong? Of course! I turned the pacing of shot sequence down. Now I could do it: From a real opponent, underspin takes longer to generate than topspin. I needed the robot to reflect the pacing of the shot in a game to enable my imagination to catch up.

I began to apply other PATT principles to Donn’s suggestion: An extremely high speed ball delivered with great rapidity by the robot might allow a fast use speed block But if I set variable placement of the ball using the oscillation of the robot to make it harder to be in position, then different actualized relationships between me and the ball would occur. If my preparation was sub optimal I could choose a defensive reduce speed shot type.

With this modification at least one aspect of judgment had to be used to select the appropriate shot type.

The robot could be set to deliver balls that fell into those areas which are commonly in between shot types. Half-long underspin (second bounce just off the end of the table) demands a choice between offense, offensive control and control type shots. If I could create an optimal set of relationships with the ball and table, an offensive shot should be attempted, if not, I would have to choose an  offensive control topspin or a control underpsin shot type.

Utilizing PATT principles I had finally moved beyond the tendency to assume that every ball emerging from the robot demanded the same shot type.

Good training should ultimately emulate match play. This suggested that using a programmable robot, the sequences of shots should emulate the types of sequences endemic to match play. Long underspin to which I would respond with topspin would be followed by repeated long topspin, demanding game-like responses. The oscillating arm could be activated more or less to require early judgment of direction. The possibilities for building judgment into constructive use of the robot were becoming apparent.


We all love robots! They give us such a great sense of comfort and accomplishment. We would like to think of robot practice as akin to taking shot practice on basketball court without an opponent present. But rehearsing table tennis shots without utilizing judgment is much more like learning to write without ever making sentences. There is no replacement for the human component in table tennis training: Predicating your judgment upon your opponent’s behavior is the basis of effective play. But with the application of principles to robot practice you can avoid becoming one of those players everyone knows; one with elegant strokes, magnificent in practice, but who just can’t compete in match play.