Quantifying Shot Quality and Predicting the Goal Probability for Football Shots

In the unpredictable world of football, primitive techniques are used to evaluate player performances, recruitment, and to build strategies for opponents. We aim to offer an improved and advanced technique of using predictive data like expected goals rather than descriptive data like shots taken and goals scored to analyse the game. Our goal is to judge teams and players on the basis of their performances instead of the results and generated predictive data for scouting and strategy formation. This was achieved by using fixed parameters on machine learning algorithms. The expected goals method gives a number between 0 and 1 for every shot taken, that number is interpreted as the probability of that shot being converted to goal. So if a shot at a particular location and at a certain angle produces an expected goal value off 0.67 then the probability of that shot to be a goal will be 67% meaning if 100 shots are taken from that same position and angle, 67 of those shots will result in goal and 33 shots will not be converted to goal. This method of using predictive data like expected goals is better than using primitive and orthodox descriptive data like total shots taken and shot on target ratio because this strips down the luck factor and just focuses on pure skill and ability. This will be beneficial while finding out players with real and hidden talents as well as analysing performances in an unbiased manner without the influence of final result.