Reconnaissance Blind Chess 4+
Johns Hopkins University Applied Physics Laboratory
Designed for iPad
Chess variant designed for research in artificial intelligence (AI).
The Reconnaissance Blind Chess (RBC) variant was invented in 2015 at JHU/APL as an experimentation platform to study artificial intelligence (AI). RBC includes imperfect information, long-term strategy, explicit observations, and very little common knowledge between opponents.
In the RBC variant, the opponents cannot see each other’s pieces but can learn about them through private, explicit sensing actions.
In unranked RBC games, the player plays with a guest account. They choose their opponent from a list of available bots and select their color (white or black). An unranked RBC game will not affect the player's rating.
In ranked RBC games, the player logs in with their RBC account credentials, and their opponent and color are chosen at random. It may take a few seconds to be matched up against an opponent. A ranked RBC game will affect the player's rating.
Reconnaissance Blind Chess (RBC) was intentionally designed to keep the game as close to traditional chess as possible while adding an appropriate amount of uncertainty and explicit sensing. RBC is like chess but includes the following core differences:
A player cannot see their opponent's pieces.
Prior to making each move, a player selects a 3×3 region of the chess board (via green outline on the relevant squares) to “sense” the pieces in that region. The player is informed of the true piece configuration within the sensed 3×3 region. The opponent is not informed about where the player sensed.
When a player captures a piece, they are informed (via red outline on the relevant square) that they made a capture, but they are not informed about what piece they captured.
When a player's piece is captured, they are informed (via red outline on the relevant square) that their piece was captured, but they are not informed about what piece captured it.
There is no notion of check or mate because neither player can be definitively aware of any check relationship.
A player wins by capturing the opponent's king or when the opponent runs out of time. Each player begins with a cumulative 15-minute clock to make all their moves.
If a player tries to move a sliding piece through an opponent's piece, the opponent's piece is captured and the moved piece is stopped where the capture occurred. The moving player is notified of the square where the piece landed (via red outline on the relevant square), and both players are notified of the capture in the manner stated above.
If a player attempts to make an illegal move, such as a pawn diagonally to an empty square, or a pawn forward to an occupied square, or a castle through an interposing piece, they are notified that the move did not succeed and their turn is over. However, castling through check is allowed because the notion of check is removed.
There is a "pass" option that a player can select if they prefer to make no move during their turn.
A player has access to reserve pieces that they can drag on and off the board to better visualize the state of the board, based on their belief of where their opponent's pieces are located.
The turn tracker informs the player when it is their turn to sense and move, and when it is the opponent's turn.
Ratings and Reviews
Data Linked to You
The following data may be collected and linked to your identity:
- Contact Info
Data Not Linked to You
The following data may be collected but it is not linked to your identity:
Privacy practices may vary, for example, based on the features you use or your age. Learn More
- Johns Hopkins University Applied Physics Laboratory
- 21.9 MB
- Requires iOS 14.0 or later.
- Requires iPadOS 14.0 or later.
- iPod touch
- Requires iOS 14.0 or later.
- Requires macOS 11.0 or later and a Mac with Apple M1 chip or later.
- Age Rating
- © 2022, Johns Hopkins University Applied Physics Laboratory, LLC