Leveraging Machine Learning Algorithms for Football Predictions and Wager Suggestions

Growing public interest in football, particularly in the wake of the recent world cup, has had an impact not just on people’s desire to watch the game, but also on the different analytics conducted in relation to it. The availability of clean, consistent data greatly enhances these analytics. This study’s primary goal is to demonstrate how various football statistics, such as goals, shots, possession, passing accuracy, and others, can be used to create an effective prediction model that will meet our primary goal of achieving over 60% test accuracy. (Given the nature of sports, it is typically difficult to quantify all the factors directly influencing the game. Therefore, although they could directly affect the outcome of the game, things like injuries sustained in the middle of play, weather, and interpersonal problems between players cannot be recorded as metrics to be worked upon. Verifying the generated winning probabilities against the odds provided by well-known bookies also made sense. The model will excel as it learns from the most recent data by dynamically obtaining the data at regular intervals through API requests. Also demonstrated is a flimsy attempt to gauge fan sentiment toward their team, which was measured and subsequently incorporated as a bias in the prediction model.