Due to the fact that sport is widespread and so popular, predicting the Points Per Game (PPG) of football players poses an interesting challenge. Hence along with the PPG, predicting the results of football matches is also a difficult task because of the number of variables which should be taken into consideration, that could not be valued or modeled quantitatively. The number of goals scored by each player has traditionally been used by many methods to predict the PPG of professional football players which would be helpful to estimate future results and evaluate the players performance. This research explores various machine learning approaches to predict the PPG of Football players by using various internal factors such as Age, Best position, Potential etc. instead of the number of goals scored by the players. In this research a model is developed that would help us to measure the PPG of the players, rather than using the actual number of goals scored and to identify best player, best club, preferred foot and relation between overall ratings and players potential. We integrated this system with a measurement of various attributes such as age, best position, nationality etc. of a football team players that are updated after each game and used to construct a model that predicts the PPG of the players, as well as to find out the best overall player, preferred foot, best club, relation between player potential and overall ratings etc. In this project I have developed three distinct feature sets for our study, which I have used on our four machine learning models, Random Forest Regressor, XGBoost Regressor, Elastic Net Regressor and Linear Regression. All the models have been evaluated on a minimal error value basis. After successful implementation of all the model we archived a minimum error of around “50.49” with the XGBoost Regressor.

Keywords: Football, PPG prediction, Machine learning, Regression models.