The Hot Stove Effect: Why AI Learns to Be a Pessimist
5 Apr 2025
Why learning algorithms are pessimistic: Hot Stove Effect shows negativity bias persists in Bayesian & sampling-based learning.
Ellipsoid Algorithms as a Tool Against Predictable Opponents
24 Jan 2025
Discover strategies to beat Follow-the-Leader and Limited History opponents in zero-sum games using algorithms like ellipsoid for prediction and counterplay.
Understanding Bias-Driven Opponent Models in Competitive Gameplay
24 Jan 2025
Learn about behavioral biases in opponents during zero-sum games, including strategies like Myopic Best Responder, Gambler's Fallacy, and more.
Ways to Counter Limited-History Opponents with Algorithmic Tools
24 Jan 2025
Algorithm 7 leverages ellipsoid prediction to beat the Limited-History Follow-the-Leader opponent in zero-sum games, minimizing losses to O(n^4 log(nr)+nr).
Future Directions for Exploiting Behavioral Biases in Games
24 Jan 2025
Future work includes exploring exploitability in probabilistic strategies, regret-minimizing opponents, and complex game structures like extensive-form games.
Broader Insights into Exploitable Strategies in Zero-Sum Games
24 Jan 2025
Generalizing strategies for behaviorally-biased opponents, this section identifies conditions for exploiting deterministic strategies to win consistently.
How Behavioral Biases Shape Gameplay Without Payoff Visibility
24 Jan 2025
Explore how behavioral biases in opponents can be exploited to win nearly every round in symmetric, zero-sum games, even with minimal information.
Methods for Decoding Opponent Actions and Optimizing Responses
24 Jan 2025
Explore predicting opponents' actions and learning best responses in zero-sum games. Learn why consistent matrices aren't enough to exploit behavioral biases.
The Key to Defeating Win-Stay, Lose-Shift Opponent Variants
24 Jan 2025
Discover strategies to beat the Win-Stay, Lose-Shift opponent variants in zero-sum games, exploiting Tie-Shift and Tie-Stay behaviors for consistent wins.