About

Palamedes is a research project focused on developing expert-level artificial intelligence for backgammon and related variants. It explores whether techniques used by strong backgammon programs can be generalized across multiple variants, and investigates new algorithms and methods to advance expert AI for these games.

Beyond strong play, Palamedes emphasizes explainability, reproducibility, and extensibility, making it suitable for researchers and competitive players alike.

Methods

Backgammon Variants

  • Greek Tavli (Portes,Plakoto,Fevga)
  • Narde
  • Backgammon, HyperGammon, Nackgammon, Takhteh
  • More variants planned soon

Evaluation

  • Neural Networks
  • Adaptive strategy on individual matches
  • Adaptive strategy on tavli matches
  • Bearoff Databases
  • Plakoto Pin Databases

Search algortihms

  • All fundamental expectimax-style search algorithms
  • Advanced chance-based search algorithms
  • New and novel search algorithms

AI Training

  • Standard self-play Reinforcement Learning (RL)
  • New approaches in self-play RL

Outputs

Palamedes (Windows)

A Windows desktop application that allows the user to play against the AI, store and replay games, and perform position analysis.

Download

HappyGammon (Android)

An Android application built on Palamedes technology, that allows the user to play against the AI in his/her device.

Google Play App page

Papers & Publications

Technical papers and research articles describing the algorithms, training methods, and evaluation results behind the Palamedes research project.

View papers

Awards & Recognition

Computer Olympiad