Rock paper scissors prediction demo

Select your move below. The neural network has already predicted your move. Feel free to play multiple rounds; the model will adapt to your playstyle over time.

How does this work?

This demo evaluates a convolutional neural network to predict a human player’s next move with up to 93% accuracy. While rock paper scissors may appear to be a game of pure chance, the human brain is notoriously flawed at generating random sequences and will ultimately form patterns (e.g. through “strategies”) that can be exploited by deep learning methods. Our proposed use of self-feature prediction approach provides unprecedented insights into designing novel loss estimation functions for deep learning architectures. Learn more about our approach in our research paper.