This piece of code predicts which choice out of three (Rock, Paper, Scissors) the human user will make with statistical significance.
When running, a UI appears, and the user may start right away. The task of the human user is to perform a series of presses on the keys "R", "P", and "S". The accuracy of computer's guesses is presented to the user and it's quite hard not to tap away a sequence that has been read by the machine. If the user taps slowly, the accuracy of the guess may decrease, but if the tapping is fast it's nigh impossible to outsmart the AI.
Thea reason that this works is that when tapping away quickly, while we may want to perform random operations, the actual keypresses are still governed by a part in our brain which means it’s a series generated by a neural network. The transfer of potential inside the human brain from one neuron to the next takes a few milliseconds, and so, when somebody is tapping several times each second there are less than one hundred neurons that can be activated in between two strokes, which limits the complexity of the model (the part of the human brain that is generating the sequence) when considering that one is tapping away a few times per second.
Having said that, it’s probably clear that taking some extra time between strokes will tend to hinder the prediction accuracy of the model that is trying to read your mind.