project proposal

< Back

Problem Statement

• We want to decide whether to kick off a project that uses readily available NFL data to apply machine learning to predict NFL game outcomes

Context

• The NFL has some good play-by-play data, and we'd like to understand the difficulty and feasibility of using this data as input to a machine learning projects.

Criteria for Success

• Provide a rough level-of-effort to predict NFL game outcomes using machine learning.

• Create a proof-of-concept to show how data can be formatted as input to a machine learning model.

• Show that the model can learn to predict NFL game outcomes better than a guess.

Scope of solution

• We don't need to predict games accurately. We just need to show that we can train a model with data that we curated to predict NFL game outcomes better than a guess.

Constraints

• Because we have 2 weeks to decide whether to green-light a project to predict NFL game outcomes, we will not try to perfect either the model or the data. We'll just see if we can get something working with a single input supervised model. We do not have access to the NFL API, which may provide better data.