Rock Paper Scissors against the computer

No. The opponent’s move is chosen BEFORE yours — right after the previous round — and the page says so under the arena. The prediction uses only the history of your past moves, so you beat it by playing less predictably.

The learning opponent keeps one model: how often each of your moves follows the previous one. The master keeps seven — move frequency, chains over your last one, two and three moves, your reaction to its own last move, and plain repetition — and each of those gets two “rotations” in case you are second-guessing it. Whichever guessed best recently makes the move; if none of them is guessing, the move is random.

Yes, but only by playing genuinely at random — that averages out to a tie on points. Any pattern (and people almost always make one: they repeat a move twice less often than chance, and they over-play whatever would have beaten the opponent’s last move) is exactly what it catches you with.

Lizard eats paper and poisons Spock; Spock smashes scissors and vaporizes rock; everything else is as in the classic game. Five moves give ten outcomes instead of three, so draws become noticeably rarer. After each round the line under the arena states what beat what.

Spock is the Vulcan salute: four fingers extended with a clear gap between middle and ring. Lizard is four fingers held tightly together with the thumb out to the side, like a puppet mouth. The line above the video shows which gesture the model sees right now, so you can settle your hand before the shoot.

Yes, and that is the main way. The move buttons are always on screen, even while the camera is open: if a gesture is not recognised the round is not lost — you can just press a move. The R, P and S keys work too (plus L and V for lizard and Spock) and the digits 1 to 5.

Settings survive a reload — variant, opponent level, match mode and your best win streak. The score and the AI’s learning start fresh, and the “Reset score and learning” button also wipes what the AI has worked out about you.