GIB: Imperfect Information in a Computationally Challenging Game

  title={GIB: Imperfect Information in a Computationally Challenging Game},
  author={Matthew L. Ginsberg},
  journal={J. Artif. Intell. Res.},
  • M. Ginsberg
  • Published 3 June 2011
  • Computer Science
  • J. Artif. Intell. Res.
This paper investigates the problems arising in the construction of a program to play the game of contract bridge. These problems include both the difficulty of solving the game's perfect information variant, and techniques needed to address the fact that bridge is not, in fact, a perfect information game. GIB, the program being described, involves five separate technical advances: partition search, the practical application of Monte Carlo techniques to realistic problems, a focus on achievable… 

Figures from this paper

Understanding the Success of Perfect Information Monte Carlo Sampling in Game Tree Search

Synthetic game trees are used to identify game properties that result in strong or weak performance for PIMC search as compared to an optimal player, and it is shown how these properties can be detected in real games and demonstrate that they do indeed appear to be good predictors of the strength of PimC search.

HyperPlay: A Solution to General Game Playing with Imperfect Information

The HyperPlay technique is described, shown how it was adapted for use with a Monte Carlo decision making process and experimental results for its performance are given.

Symmetries and search in trick-taking card games

This dissertation constructs symmetry-based search extensions for efficiently solving perfect information game states, for use in a Monte Carlo-based high-performance computer Skat player, and extends these symmetries to construct precomputed endgame lookup tables, reducing search time substantially.

General Game Playing with Imperfect Information

This article considers the problems of working with imperfect information and introduces a technique, called HyperPlay, for efficiently sampling very large information sets, and presents a formalism together with pseudo code so that others may implement it.

On the Complexity of Trick-Taking Card Games

A general class of perfect information two-player trick-taking card games dealing with arbitrary numbers of hands, suits, and suit lengths is defined and a proof of PSPACE-completeness for a fragment with bounded number of hands is proved through a reduction from Generalized Geography.

Ensemble Determinization in Monte Carlo Tree Search for the Imperfect Information Card Game Magic: The Gathering

This paper examines the use of Monte Carlo tree search (MCTS) for a variant of one of the most popular and profitable games in the world: the card game Magic: The Gathering, and examines the effect of utilizing various pruning strategies in order to increase the information gained from each determinization.

Search in Imperfect Information Games Using Online Monte Carlo Counterfactual Regret Minimization

This paper presents Online Outcome Sampling (OOS), the first imperfect information search algorithm that is guaranteed to converge to an equilibrium strategy in two-player zero-sum games and shows that unlike with Information Set Monte Carlo Tree Search (ISMCTS), the exploitability of the strategies produced by OOS decreases as the amount of search time increases.

Monte-Carlo Tree Search for the Game of "7 Wonders"

This work studies Monte-Carlo Tree Search on the game of "7 Wonders", a more complex game that gathers together several known challenging properties, such as hidden information, multi-player and stochasticity, and shows that many known results still hold.

Determinization and information set Monte Carlo Tree Search for the card game Dou Di Zhu

A novel variant of MCTS that operates directly on trees of information sets is introduced and it is shown that this algorithm performs well in precisely those situations where determinization using random deals performs poorly.

On Monte-Carlo tree search for deterministic games with alternate moves and complete information

It is proved that the step by step optimal algorithm built here is not globally optimal and that it always converges in a finite number of steps, even if the a priori the authors have on the game is completely irrelevant.



Finding Optimal Strategies for Imperfect Information Games

These algorithms theoretically and experimentally are compared using both simple game trees and a large database of problems from the game of Bridge, showing that the new algorithms both out-perform Monte-carlo sampling, with the superiority of payoff-reduction minimaxing being especially marked.

Generating and Solving Imperfect Information Games

Gala is described an implemented system that provides the user with a very natural and expressive language for describing games and implements the first practical algorithm for finding optimal randomized strategies in two player imperfect information competitive games.

Search in Games with Incomplete Information: A Case Study Using Bridge Card Play

Asymptotic Properties of Minimax Trees and Game-Searching Procedures

  • J. Pearl
  • Computer Science
    Artif. Intell.
  • 1980
It is shown that a game with WIN-LOSS terminals can be solved by examining, on the average, O [(d) h 2 ] terminal positions if positions if P 0 ≠ P∗ and O [(P∗ (1 − P ∗) ) h ] positionsif P 0 = P∷, the former performance being optimal for all search algorithms.

A Review of Game-Tree Pruning

  • T. Marsland
  • Computer Science
    J. Int. Comput. Games Assoc.
  • 1986
These essential parts of game-tree searching and pruning are reviewed here, and the performance of refinements, such as aspiration and principal variation search, and aids like transposition and history tables are compared.

Conspiracy Numbers for Min-Max Search

Exploiting Graph Properties of Game Trees

This paper exploits the graph properties of the search space to reduce the size of trees built in practice by at least 25%, proving that there is more room for improvement of current algorithms.

Using Patterns and Plans in Chess

Alpha-Beta Pruning Under Partial Orders

This work generalizes to game trees with position values taken from a partially ordered set, and proves necessary and sufficient conditions for alpha-beta pruning to be valid, and shows that the resulting technique leads to substantial improvements in the speed of algorithms dealing with card play in contract bridge.

Opponent Modeling in Poker

Loki is described and evaluated, a poker program capable of observing its opponents, constructing opponent models and dynamically adapting its play to best exploit patterns in the opponents' play.