Hi All,

I have recently posted the following article:

https://towardsdatascience.com/ai-in-video-games-improving-decision-making-in-league-of-legends-using-real-match-statistics-and-29ebc149b0d0

In this project, I took real match statistics and attempted to model it as an MDP and then apply reinforcement learning in order to find the next best play for long term .

I have also posted all of my working in three parts on Kaggle:

https://www.kaggle.com/osbornep/lol-ai-model-part-1-initial-eda-and-first-mdp

https://www.kaggle.com/osbornep/lol-ai-model-part-2-redesign-mdp-with-gold-diff

https://www.kaggle.com/osbornep/lol-ai-model-part-3-final-output

I will be trying to share with the subreddit to see how they feel about this kind of idea but would appreciate your thoughts on whether this makes sense and if my model is applied correctly.

Thanks



Source link
thanks you RSS link
( https://www.reddit.com/r//comments/90w1t3/p_using__in_league_of_legends_to_improve_my/)

LEAVE A REPLY

Please enter your comment!
Please enter your name here