Application of machine learning to help AI to play Hearthstone

This paper presents a solution, which was developed as a part of the competition AAIA'17 Data Mining Challenge: Helping AI to Play Hearthstone. The goal of the competition was to predict the probability of AI player win in different intra-game states of Hearthstone game (online computer game with cards). This solution got the third place at the final leaderboard. The paper describes models and local validation approach, which was very useful for models development without overfitting.