Forecasting 2018 NFL Rookie LB Success: 3-Year Model
In this series of articles for the past few seasons, we have been using machine learning and statistical models to forecast the future success of rookie running backs and wide receivers. This season, I am diving into individual defensive players (IDP) for the first time, so I will be applying the same toolkit to the most important individual defensive position: Linebackers.
If you play IDP, then you will also want to read Mike Woellert's article on these rookie linebackers to get his expert perspective on their abilities, as well as the opportunity available for them to contribute to their new teams.
By the way, if you’re in a dynasty league, you should definitely check out all the content from our friends at Dynasty League Football. They also have more information on the rookies I talk about below.
Forecasting Linebacker Success
We will be following the same approach that has worked well for offensive players: using an ensemble model based on the predictions of a statistical model and a machine learning (ML) model.
Both models take into account draft capital, college production, and athleticism. The machine learning model uses support vector machines (SVM), which mix all of these inputs together in a sophisticated manner to make a prediction.
The statistical model, while in some ways less sophisticated, is much easier to interpret. It considers only the following factors: Draft pick, age, market share of tackles in their final season of college, market share of forced fumbles in their final season, and weight. As usual, draft pick is by far the most important factor. Furthermore, the first three of these parameters—draft capital, age, market share of tackles—are fairly accurate predictors by themselves (without forced fumbles or weight).
The goal of both models is to predict the odds of having a top-24 season in one of their first three years in the NFL.
Below you will find a table with complete forecasts and tiered rankings for the 2018 rookie linebackers, as well as my breakdowns of a few names that stood out...