GDC is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

View, browse and sort the ever-growing list of sessions by day, pass type, topic, and format. With this Session Viewer, you can view GDC 2023 session details and speakers, and share your favorites via social media. You will be able to build your schedule and access it during the show via export or Mobile App, once live. Sessions do fill up and seating is first come, first serve, so arrive early to sessions that you would like to attend.

Machine Learning Summit: Simulating Teamfight Tactics Using Deep Learning for Fast Reinforcement Learning AI Training

Ran Cao  (Staff Data Scientist, Riot Games)

Location: Room 2010, West Hall

Date: Tuesday, March 21

Time: 3:50 pm - 4:20 pm

Pass Type: All Access Pass, Summits Pass

Topic: Programming

Format: Lecture

Vault Recording: Video

Audience Level: All

In 2019, Riot Games launched Teamfight Tactics (TFT) on an expedited timeline. Since then, the TFT game team has been heavily focused on launching new features and content to improve the live game experience.

In this talk, Ran Cao, Staff Data Scientist at Riot Games, will present a lightweight path their scientists used to make progress building reinforcement learning agents that learned to play TFT. They built their own version of the game outside of the game engine that leveraged a neural network to predict outcomes rather than fully simulating the game. With this version of the simulator, they are able to minimize the amount of resources required from the game team and still be able to train agents to play TFT at a high skill level and provide gameplay insights.

In addition, they were able to make changes to the game in the simulator and test hypotheses with AI that would have been extremely costly to test in the real game.


Attendees will gain insight into an alternative path for demonstrating progression on complex tasks like reinforcement learning without impacting game teams' roadmaps, as well as examples into new types of visualizations and tools that can be leveraged by design teams.

Intended Audience

This is for AI/Gameplay programmers and game designers who are interested in the application and implementation of reinforcement learning for game AI, as well as ML for complex simulation. Basic understanding of machine learning is recommended.