nipunbayas / pmpi

Stony Brook University- AI Course Project

Player Match Performance Index

Stony Brook University - CSE 537 - AI Course Project

We are using the dataset available at https://www.kaggle.com/hugomathien/soccer

In this project, we are predicting the likelihood of a player to score in a particular match. In order to make the predict, we have used a Bayesian Network which is based on the following parameters:

• Team Strength
• Team Form
• Individual Player Rating
• Individual Player Form

Each of these factors is based on the following individual factors:

• Team Factors

  • Number of goals scored in the past few matches
  • Number of wins, draws and losses for that particular team in the last few matches
  • Total points scored by a particular team in the past few seasons
  • Ratings (strength) of players playing in that team

• Player Factors

  • Form of the player's team (based on the past 5 matches)
  • Position of that players i.e. Goalkeeper, Defender, Midfielder or Forward
  • Form of that player in the past 5 matches by computing the number of goals, assist and shots taken by the player
  • Overall rating of that player based on his attributes

After feeding these values to the Bayesian Network, we come up with the top 5 players who are likely to score in a selected match. We then check the results of our prediction with the actual result of that match.

We have tested our algorithm on the English Premier League, but this can be easily extended to any season of any league.