Creating an ML Model to Predict the Winners of Mirror Matches in Super Smash Bros. Ultimate

Every once in a blue moon, when my friend Travis and I end up with several hours of time to kill while together and also in the presence of a Nintendo Switch, we run the entire gamut of characters in the game back to back. This usually takes us somewhere between 5 and 6 hours depending on the breaks that we take, and at this point, after all DLC characters have been released, we play 83 matches of smash during the ordeal. By the end we're usually completely sick of the game.

We recently found ourselves in a situation where we had the opportunity to do this, and it had been long enough since the last time that we did, so we decided to give it another go, and this time I decided to keep track of some stats and see what I could do with the data, giving rise to this project.

Throughout the project we will construct both k-nearest and logistic regression models to predict the outcomes of each match. First we will use a limited dataset, and then join that dataset together with a large set of frame data for each fighter’s moves and create new, larger models based off this new data, and see how the models compare.