On the Limitations of Elo: Real-World Games Are Transitive, Not Additive

Explore the limitations of Elo rating systems in real-world competitive games during this 28-minute GERAD Research Center talk. Delve into the challenges of identifying the strength of transitive components in games like chess, go, and StarCraft II. Discover how Elo models can fail to extract transitive components even in elementary transitive games. Learn about a proposed extension to the Elo score, introducing a disc ranking system that assigns players both skill and consistency scores. Examine the empirical validation of this new approach using payoff matrices from real-world games played by bots and humans.