finding square pegs for square holes: identifying player ... · blocked shots focus on data...
TRANSCRIPT
@OptaPro
#OptaProforum
Finding square pegs for square holes:
identifying player types for scouting
Footballers have a very particular set of skills, skills they have acquired over
a very long career.
Assess these skills and classify footballers into different types based on
their statistical ‘traits’.
Goal of analysis
Image Credit: LiverpoolFC.com
Dataset & method
Defensive Passing & Creativity
Shots (Open-play)
Other
Tackles Passes Shots Inside the Box
Aerial Duels
Challenges Chances Created (Open-play)
Shots Outside the Box
Dribbles
Interceptions Crossing (Open-play)
Fouled
Fouls Long Balls
Clearances Through-balls
Blocked Shots
Focus on data averaged over a season (data averaged to per 90 minutes played). Data from Premier League, Bundesliga and La Liga 2012/13 & 2013/14.
Use Principal Component Analysis to combine the data and construct underlying structure. Then run Cluster Analysis to identify different player types based on their traits.
Principal components
Attack
Defence
Involvement Heading
Cluster analysis: twelve cluster solution problem
Creative midfielders & Dani Alves…
Cluster analysis: twelve cluster solution
Cluster analysis: forwards
Creative shooters Combine shots with creativity and dribbling.
Penalty-box strikers Biased towards shots in the box. Less involved in play. Limited creativity.
Target-men Heavily involved in aerial duels. Lower shot volume. Limited creativity.
63% of players stay within the same cluster
Cluster analysis: midfielders
Attacking midfielders Mix creativity with shots. Limited defensive contribution.
Through-ball merchants Very creative, love a through-ball. Limited defensive contribution.
Disruptors Big defensive contribution, limited playmaking.
Deep-lying playmakers Playmaking from deep (long balls and chance creation). More limited defensive contribution.
Creative midfielders Playmaking duties. Mainly long shots. Average defensive contribution.
Defensive controllers Big defensive contribution, combined with playmaking.
Direct attackers Combination of dribbling and shots inside the area. Limited creativity.
70% of players stay within the same cluster
Cluster analysis: defenders
Attack-minded fullbacks Biased towards attacking contribution (chance creation, crossing, dribbles).
Penalty-box defenders Defensive contribution biased towards clearances and blocked shots.
Proactive defenders Defensive contribution biased towards tackles, interceptions, fouls.
Two-way fullbacks Strong defensive numbers, plus attacking contribution.
Hybrid defenders Mix clearances, blocked shots and interceptions. Relatively fewer tackles.
71% of players stay within the same cluster
Square pegs for square holes
Forwards
Creative Shooters
Target-men Penalty-box strikers
Defenders
Two-way fullbacks
Proactive defenders
Penalty-box defenders
Attacking fullbacks
Hybrid defenders
Midfielders
Attacking midfielders
Disruptors Deep-lying playmakers
Defensive controllers
Direct attackers Through-ball merchants
Creative midfielders
Able to identify player types and create statistical baselines for performance comparisons. Can aid player scouting by helping to find the “right fit”. Identifies rare skill sets (potentially enhances value to team). Increased number of leagues and players would expand potential of the system. Can be extended to more complex datasets.
Thank you
Will Gurpinar-Morgan
2+2=11
(http://2plus2equals11.wordpress.com/)
@WillTGM & [email protected]
#OptaProForum