Join our 2019 NeurIPS Competition!

A Challenge for Making Optimal Decisions Under Uncertainty

The tournament for our NeurIPS competition is over. Come hear about it at NeurIPS 2019 and stay tuned for upcoming ways to play a bot.


Many of the favorite studied games in artificial intelligence (AI) such as checkers, chess, and Go lack something that is extremely common and critical in real-life decision making: uncertainty.

We are hosting a competition with a simple but powerful twist on what may be considered the most classic game in AI history, chess. Reconnaissance blind chess (RBC) is like chess except a player cannot see where her opponent's pieces are a priori. Rather, she learns partial information about them through chosen sensing actions and the results of moves.

In comparison to poker, which seems to be the most popularly studied game of imperfect information, RBC includes a critical component of long-term planning. Compared to phantom games like Kriegspiel, in RBC players have much more ability to manage their uncertainty, which we believe makes the game more interesting from an AI perspective and more realistic for most scenarios; players are not completely blind, but rather, metaphorically, they simply cannot look everywhere at once.


Ryan W. Gardner
JHU Applied Physics Laboratory
Casey Richardson
JHU Applied Physics Laboratory
Corey Lowman
JHU Applied Physics Laboratory
Ashley J. Llorens
JHU Applied Physics Laboratory
Andrew Newman
JHU Applied Physics Laboratory
Arec Jamgochian
Stanford University
Jared Markowitz
JHU Applied Physics Laboratory
Nathan Drenkow
JHU Applied Physics Laboratory
William Li
JHU Applied Physics Laboratory
Todd W. Neller
Gettysburg College
Raman Arora
Johns Hopkins University
Bo Li
University of Illinois
Mykel J. Kochenderfer
Stanford University
Ritchie Lee
Stinger Ghaffarian Technologies: SGT