So far, there are only table soccer games which are controlled by a PLC, an industrial controller. They only detect the ball and then shoot fast and powerfully. So they only want to hit the ball; they have no tactics, no sensitivity. Our table soccer, on the other hand, builds its knowledge by learning
how to play from the ground up. It is done via training in competition mode. The camera records the playground. This is the same way the human eye works. So the image gets transferred to the brain of the table soccer AI, gets processed, and the brain uses this know-how to make a decision how to make and
how to score the goal and which player to move. First the machine learned how to play. Now we want to win! That’s why we send it to the training camp. In training mode, the table soccer AI learns by chance. It actually remembers which of its decisions lead to goals. While doing this, its neural network is trained by rewarding based on good and bad behavior. Good behavior means that the table soccer AI made actually a goal; and bad behavior means that the human opponent of the
table soccer made a goal. We will never stop working on winning. But the machine itself will be already our first victory.