Modelling Players Prices on Football Index

Objective

We wanted to know whether players prices are rational on the index and whether they can be explained by certain variables. The results can then help in 2 ways:

  • Knowing the key variables that drive player prices on Football Index.
  • Estimating each player's price based on the variables to help identify underpriced/ overvalued players.

Methodology

We collected over 50 variables on every player on the Index where possible including:

  • Performance Buzz Stats such as their PB average.
  • Player Stats such as goals, assists, passes etc.
  • Dividends (since PB introduction) ie Media Buzz and Performance Buzz
  • Player Characteristics such as Age, Estimated Player Market Value, Nationality and Position.
  • Club

We then ran the variables through a mixed model to estimate each players price given each players current price and their variables.

Results

We found the following variables to be statistically significant (at a 10% significance level), they are displayed below in order of statistical significance:

  • Media Buzz Dividends
  • Estimated Player Market Value
  • Performance Buzz Dividends
  • Goals per 90 minutes
  • Performance Buzz Averages
  • Transfer Linked?
  • Age
  • Shots per 90 Minutes
  • Tackles per 90 Minutes by Position
  • Penalties Taken
  • Shots per 90 by Position

We had to drop some of the players from the model due to missing variables (e.g. if they haven't played this season). We will be releasing on the Members Blog each player's modelled price and those whose differ most from their actual price (hint: Milner is pretty overpriced) at 7 PM this evening.