Skip to content

junhwanjang/whoscored

Repository files navigation

Football player model

  • We created two models using game data in each league. Frist model is to check what factor is critical for player's rating. As using the model, it can predict their rating based on a single game data. Second model is that finding better position based on training data. The second model is based on about 2,500 players having a single position. Then, it will be used for helping other players having multi positions find their better position.
  • Parsed Website : www.whoscored.com
  • Sample data : Game data in 2015-2016 season
    • League : Bundesliga(Germany), LaLiga(Spain), EPL(UK), Ligue1(France), SerieA(Italy), Eredivise(Netherland), RPL(Russia)
    • Position(Class) : FW, MF, DF, GK
    • Player : Football Players having more 270 minutes on appearance.

1. Crawling game data

https://github.com/junhwanjang/whoscored/blob/master/1_Crawling%20Players'%20data.ipynb

2. Player rating model

  • Comment : I think that if we have more past season data, I am sure we can get some significant insight about changes of main factors on each position and on the overall football strategy/style.

2-1) FW

https://github.com/junhwanjang/whoscored/blob/master/2_Forward%20OLS%20analysis.ipynb

2-2) MF

https://github.com/junhwanjang/whoscored/blob/master/3_Midfielder%20OLS%20analysis.ipynb

2-3) DF

https://github.com/junhwanjang/whoscored/blob/master/4_Defender%20OLS%20analysis.ipynb

2-4) GK

https://github.com/junhwanjang/whoscored/blob/master/5_Goalkeeper%20OLS%20analysis.ipynb

3. Position recommend model

  • Train Data : Football players having single position
  • Test Data : Football players having multi positions
  • Comment : It could help players having multi positions find their better position. The model will be used for the game industry because sport itself is a kind of game.

https://github.com/junhwanjang/whoscored/blob/master/6_Position%20Classification%20model.ipynb