This model aims to use college football statistics to predict fantasy football success for a college quarterback prospect to the NFL. To do this I have created two models. One which predicts fantasy points per game (PPR). One predicts NFL draft position using a neural network. Last year I used a linear regression model. Not every metric has a linear relationship to the input and output variables, which is where linear regression comes up short. Neural Networks are the computerized representation of the human brain that finds complex relationships between input and output nodes to create weights that help predict the desired outcome.

The Results 

The data was trained on 653 QBs in the classes of 2017-2022 and the prediction was on 318 QBs in the classes of 2023-2025. The sum of squared error for the fantasy points per game model is 0.3616 and the draft position model is 0.5173. The neural network is attempting to minimize the sum of squared error as an objective function within the model, which minimize variation in the prediction from the actual result. The input variables are: 

  1. Power 5
  2. Cumulative Pass EPA: Increments of 100 pass attempts between 100-1000
  3. Cumulative Rush Yards Over Expectation: Increments of 10 rush attempts between 10-200
  4. QBR: Years 1-5 out of High School
  5. Games Played: Years 1-5 out of High School 
  6. Cumulative Sack EPA Lost: Increments of 100 pass attempts between 100-1000
  7. Cumulative Interception EPA Lost: Increments of 100 pass attempts between 100-1000

The Prediction

Below I have identified the Top 60 QBs in the model (Sorted by Fantasy & Draft QB Index) for the class of 2017-2025 and the Top 40 QBs in the model for the class of 2023-2025. Some of the most influential variables for the fantasy points per game model are cumulative pass EPA at 400 attempts, cumulative rush yards over expectation at 40 attempts, sack EPA lost at 100 attempts, and games played in year two out of high school. Some of the most influential variables for the draft position model are games played in year one, cumulative pass EPA at 900 attempts, QBR Year 1, QBR Year 3, cumulative rush yards over expectation at 200 attempts, sack EPA lost at 300 attempts, interception EPA lost at 800 attempts.

A few takeaways: 

  • The Bryce Young Discourse: Young is in the 63rd percentile in cumulative pass EPA by 400 attempts, 47th percentile in cumulative rush yards over expectation by 40 attempts, and 57th percentile in sack EPA lost on his first 100 attempts. CJ Stroud is in the 92nd percentile cumulative pass EPA by 400 attempts and Justin Fields is in the 80th percentile. Although he is praised for creating, he also sacrifices expected points through sacks much worse than other top QBs. Chicago should gain assets by trading the #1 overall pick and improving their pass catchers. For rookie drafts, I would be trading back to get Stroud + a veteran RB, a later first, and a veteran QB, or trading up for Justin Herbert. 
  • The 2024 QB class is better than the 2023 QB class at the top with Caleb Williams and Drake Maye. 
  • Michael Pratt has garnered attention as a possible NFL QB. His metrics have been steadily improving and are currently sitting around other day two and three prospects. I have been in on Grayson McCall as a potential NFL QB, but I will pivot that energy to Michael Pratt. 
  • Where is Quinn Ewers? Quinn Ewers was in the 70th percentile in cumulative pass EPA by 100 attempts but has fallen to the fifth percentile cumulative pass EPA. Ewers could reset his hot start being fully healthy in 2023.
  • Don’t draft Will Levis, he is in the 43rd percentile in cumulative pass EPA at 400 attempts.

Get Our Newsletter

Get notified periodic notifications about our content and future subscription deals.

You May Also Like

Clustering 2023 Rookie Wide Receivers based off their College Playstyle Metrics

Analyzing how a WR’s playstyle will be successful in the NFL is a fun exercise and our TIm Alb breaks it down into eight clusters. #Campus2Canton #FootballAnalytics #2023NFLDraft

2023 RB Model – Using Neural Networks

Using a neural network, @EdnaBeastMode dives into the metrics that can predict NFL success in college RBs. #2023NFLDraft #Campus2Canton

The 411: American Athletic Conference (AAC)

Hannah Page, @HanaLei91, is back for another year of the 411 Series. First up – four players, one coach, and one team from the AAC to watch in 2022! #Campus2Canton #CollegeFootball #AACFootball