Usage Scenario 4

Advanced analysis

Group players into playing-style archetypes with K-Means and summarize the value prediction model for a candidate profile.

Cluster projection uses PCA coordinates; prediction output reports estimated value and feature contribution weights.

Clusters

Mapped players

17450

Estimated value

€110,692,369

R2 score

0.971

Cluster control

Choose the number of playing-style groups

K-Means PCA map

Playing-style clusters in two dimensions

Cluster labels

Centroid summaries

Ball-Playing Defenders

4504 players

pace: 67.3, shooting: 44.7, passing: 56.9, dribbling: 60.8, defending: 62.1, physic: 67.8

All-Rounders

3318 players

pace: 68.4, shooting: 62.2, passing: 68.9, dribbling: 70.9, defending: 65.6, physic: 71.4

Pacey Attackers

3585 players

pace: 77.1, shooting: 66.8, passing: 63.5, dribbling: 71.3, defending: 37.2, physic: 63.6

Lightweight Attackers

3304 players

pace: 69.3, shooting: 55.1, passing: 51.6, dribbling: 60.7, defending: 32.2, physic: 53.6

Traditional Defenders

2739 players

pace: 57.3, shooting: 31.7, passing: 44.0, dribbling: 47.2, defending: 61.1, physic: 67.2

€110,692,369

Value prediction output

Ridge regression is trained on FIFA 22 outfield players and log1p(value_eur), matching the Predict value notebook. Features are overall, potential, age, pace, shooting, passing, dribbling, defending, physic. Model engine: sklearn; rows used: 17081. wage_eur is accepted for backward compatibility but is not used by this notebook model.

0.971

MAE

€654,442

Test rows

3417

Feature weights

Model contribution direction and magnitude

overall

1.22

age

0.45

potential

0.09

shooting

0.03

pace

0.02

defending

0.01

Feature	Weight
overall	1.22
age	0.45
potential	0.09
shooting	0.03
pace	0.02
defending	0.01
physic	0.01
dribbling	0.01
passing	0.00