Riot Games uses deep reinforcement learning to help game designers analyze and test game balance in Legends of Runeterra before content is released to players, giving designers a tool to predict whether new cards or mechanics will be overpowered or underpowered.

Riot trained a reinforcement learning agent that plays Legends of Runeterra against itself at scale to discover which decks and cards are consistently dominant or weak, producing balance scores that became part of game designers' key performance indicators. The agent's predicted balance scores were found to match design intuition and correctly predicted which decks were strongest, with directional alignment confirmed against live player data. This was Riot's first major productionized use of deep reinforcement learning for game balance and has since been extended to additional game titles.