College football is more popular than ever, with games in 2023 broadcast on an ESPN platform averaging 1.7 million viewers per game. Further, the introduction of NIL deals and the expansion of legal sports betting across the US has expanded the amount of money in the sport. In particular, the average college football program has a revenue of $31.9 million per year, and an estimated $8 billion per year is bet on the sport (at least legally).

Motivated by the above, we want to develop a predictive model(s) for the outcomes of future college football games. We develop models to predict many of the core statistics pertaining to the outcome of a given game, all the way from a given team's rushing yards or penalties up to the final score and the winner. Our best performing model is based on an LSTM network and predicted the final scores in a game with a mean absolute error of 3.45 points, and correctly predicted the winner of the game 90.5% of the time.