Funny you should mention an unreliable tic-tac-toe, I've been reading the book "Racing the Beam" and it mentions the 1956 relay-logic computer "Relay Moe", whic

Funny you should mention an unreliable tic-tac-toe, I've been reading the book "Racing the Beam" and it mentions the 1956 relay-logic computer "Relay Moe", which, despite being constructed of little more than 90 relays and a spinning drum, touted a "variable intelligence" which allowed the operator to choose how often the machine makes a mistake, and rotates the board between games so its repetitive strategy is not immediately obvious. Quoting the book (p39): > Playing against a perfect tic-tac-toe competitor, and playing well, can only ever result in a draw. This may be computationally impressive the first time, but it isn't very fun. Opponents, computer or human, become interesting when they make mistakes--or more accurately, when it becomes clear that they might make mistakes under certain circumstances. Such mistakes highlight weaknesses, which players can exploit as part of a strategy. The book goes onto explain how Pong on the Atari 2600 achieves a similar level of unreliability, since it would be trivial for the computer to follow the y-position of the ball and never lose. Every 8 frames, the computer neglects to track the ball, drifting behind the position, but recovering its position when the ball ricochets off the top or bottom of the screen. This way the "AI" of Pong occasionally makes mistakes, in order not to be too harsh to the human's ego. Thorough breakdown of Relay-Moe's logic with photos is here: (3.5MB PDF) https://www.vintagecomputer.net/cisc367/Radio%20Electronics%20Dec%201956%20Relay%20Moe%20Plays%20Tic-tac-toe.pdf https://www.vintagecomputer.net/cisc367/Radio%20Electronics%...