League of Legends AI: Bots Learn to Play Like Pros

League of Legends AI bots are getting a major reboot

Riot Games is planning to overhaul the cooperative versus artificial intelligence (AI) mode to look more like a normal match in League of Legends. The aim is for bots to perform actions such as jungling, ganking, and killing dragons, while having a general AI for the team bots to work together. The ultimate goal is to allow bots to react to changes in the metagame and learn from them.

In a lengthy blog post, technical product manager Darcy Ludington and bot team technical lead Emmett Coakley outlined how the League team intends to improve bot systems to make them more useful not only in existing bot-based game modes, but also in future AI-powered game modes. The biggest hidden goal of the cooperative versus AI overhaul seems to be to provide players with a significantly expanded way to learn and improve their League skills, which is less stressful and high-stakes than player versus player (PvP) modes.

Currently, bots do not perform basic game-related tasks, and the cooperative versus AI mode does not resemble the League game played by modern players. Expanding the pool of characters that bots truly know how to play, making them play as a team, and scaling them to match player skill levels are small steps in that direction.

“Our data shows that no matter how many cooperative versus AI games a new player played before queuing up for PvP, those games would not improve a new player’s chances of winning.” – Coakley and Ludington

The ultimate goal is for game designers to be able to modify bots. Ludington and Coakley say they want to “build our bot technology to be scalable, maintainable, and extensible so that designers have fun levers to pull to delight players.”

The aim is for the beta of the new bots to be available in the public League beta environment for two weeks in the second half of this year—sometime between July and December. This test will be only the first round of testing, as the team hopes to “get a few rounds of feedback before launching.”

Source: PC Gamer.