NFL meets Premier League

Summary of process:

Used and enhanced a pre-existing database of NFL statistics for over 1,700 players across 22 years of teams to develop a predictive model relating physical attributes of players and success in the regular season . Created a database for the 2025/2026 premier league team’s players for age, weight, height and FIFA pace ratings. Calculated averages for each category for each team on the NFL and the premier league. Used a regression model to approximate 40 yard dash times from FIFA pace ratings. Using machine learning it was determined that the best model for bridging American football and premier league teams was boosted ensemble using all four variables.

Issues and resolutions:

Prompting AI to collect and analyse the data to form the premier league dataset provided issues. Developed prompting skills and communication with AI by being more specific. FIFA pace ratings were .png files which AI had trouble extracting. Manually went through the data and calculated an average. Combining the combine data into a large CSV file for NFL. There was no direct access to the premier league average speeds. Formed a regression model to bridge FIFA pace ratings and 40- yard dash times. Initially, actual wins vs. predicted wins didn’t fit the trend exactly due to the average weight of NFL players being much higher than premier league players. We took into account that the opponents of the premier league players would be of a similar weight eliminating this issue by shifting the mean.

Applications:

This model could be used to bridge other sports to see how they would compare to playing American football. Can be used to make preseason rankings when available data is limited.