← All work
Analytics & ML · 2023

What drives a Premier League player's transfer value?

R² 0.684of market value explained
Role

Solo · replication

Year

2023

Type

Analytics & ML

Tools

Regression, Statistics

The answer

Look for productive players with heavy minutes at lower-ranked clubs; discipline and physique barely register.

Case files on GitHub ↗
Age and club rank drag value down; goals and minutes push it up
Age: −12.32Age−12.32Club rank: −9.93Club rank−9.93Average goals: 4.92Average goals4.92Average playing time: 3.89Average playing time3.89Average appearance time: 1.92Average appearance time1.92Average clean sheets: 1.75Average clean sheets1.75Height: 1.66Height1.66Penalty goals: −1.43Penalty goals−1.43Weight: 1.12Weight1.12Yellow cards: −0.91Yellow cards−0.91Average assists: 0.8Average assists0.8Position: −0.47Position−0.47Penalty misses: 0.39Penalty misses0.39Player rank: 0.36Player rank0.36Red cards: −0.08Red cards−0.08

t-values from the 15-variable regression. Red: significant at p < 0.001. Replication of Li (2020).

Source: tables/table4a_regression_coefficients.png

0.684R², 15 variables
324players
−12.32t-value for age

01The problem

Which player attributes actually move transfer value, and which don't?

Solo research project, Semester 2, 2023 · A replication of Li (2020)

02What I did

  • Rebuilt the published model, as a replication of Li (2020).
  • Ran correlations and a multiple regression with 15 predictors.
  • Checked validity with an F-test, VIF and Durbin-Watson.

03Results

Model fit
Players324 (2018–19 Premier League season)
Predictors15
R² / adjusted R²0.684 / 0.668
F-statistic44.237 (p < 0.001)
Max VIF4.73
Durbin-Watson1.82

Source: report; tables/table1_data_sample.png; report; tables/table4b_model_fit.png; tables/table4a_regression_coefficients.png

04Files

Case files on GitHub ↗
Next project

ESG sentiment classifier

2 of 5 →