The game recognises your emotion from your facial expression using a neural network trained. This might sound complicated, but it doesn't require a heap of maths, code, or compute. It was trained on 500 images of each expression using a sample of AffectNet, a labelled dataset of facial expressions (some of which are pretty noisy). These were trained in a few lines of code using fastai using material from the first chapter of the fastai book or the first lesson of the fastai course. The training was done using free compute using a Kaggle notebook, and is hosted freely as a huggingface space.