Algorithms & Models
Our solutions use statistical data from present and past years, along with machine learning practices to develop sports data analysis algorithms. Our models have been created by our team while attending Columbia University and are built using software tools for machine learning, data visualization and data manipulation such as Scikit-Learn, Pandas, TensorFlow.
The sides prediction model uses historical data from the past 10 years. The model is an algorithm composed of our own metrics that we have deemed most influential to the outcome of a football game. We are most proud of our QB metric, which we have created from scratch that measures how a specific QB performance will impact the outcome of the game. The model is coded through python and uses machine learning logistic/linear regression to predict the outcomes.
The totals prediction model uses historical data from the past 10 years. The model is an algorithm composed of our own metrics that we have deemed most influential to the scoring of total points in a football game. Just like the sides model, we are most proud of our QB metric, which we have created from scratch that measures how a specific QB performance will impact the scoring of the game. The model is coded through python and uses machine learning linear regression to predict the final scores.
The player model uses historical data of a particular player from the past few seasons. Given the game matchup and location of the game, predictions are made on rushing, receiving and defensive plays.