Judging in diving is supposed to be as objective as possible and fairly account for all phases of a dive: start, take-off, flight, and entry. However, in practice, it is a commonly held belief among competitive divers that the certain phases of the dive matter more than others when scoring. My project seeks to answer whether or not that is true. I use machine learning to attempt to mimic a human judge’s scoring style, and measure how much each phase is “weighted” when scoring. Dives are broken down to their component phases and feature vectors are extracted using pre-trained spatiotemporal 3D Convolutional Neural Networks (C3D). Feature vectors from the multiple phases are weighted and multiplexed together to form the input to a fully-connected neural network which outputs a predicted score. Training of the neural network and regression of the weights are performed alternately in multiple epochs. The result is a weight vector that indicates the relative importance of each diving phase in the scoring. I find that the start and end phases are consistently weighted higher in scoring than the middle phases, which may be due to the cognitive bias that humans pay more attention to the beginning and end parts of a series than the middle. This conclusion suggests that we may need to rethink the scoring system in diving such that it more fairly takes into account all phases of the dive.