Regression based machine learning on the FIFA Ultimate Team transfer market

Abstract

Ultimate Team is a highly popular game mode in the FIFA video game series developed by EA Sports. In Ultimate Team, players can buy and sell items based on real players on the transfer market with an in game currency. To combat buying and selling the currency with real money, the items have a price range set on the transfer market so that each item can only be sold for a reasonable price based on the item’s abilities. This thesis demonstrates how machine learning regression models could be used to predict the prices of new items entering the game, so that the price ranges could be set more accurately to improve player experience. As the predictions weren’t good enough to improve the current situation, the thesis goes through what are biggest issues in making the predictions.