Crazy Stone Deep Learning -The First Edition-

Crazy Stone Deep Learning -The First Edition- is a Go program that uses deep convolutional neural networks in place of a pattern database, combined with Monte Carlo Tree Search. It offers 20 play levels, from 13k to 7d, for each of its three board sizes. The lower levels are intended for average players.

Game modes

You can play Human vs Computer, Human vs Human on a shared device, or Computer vs Computer. Handicap games and variable Komi options are available, as are timed Byoyomi games. In timed games, you cannot select the computer level.

Features

• Boards and levels: 20 levels for each board size: 9×9, 13×13 and 19×19.
• Rules and timing: Japanese or Chinese rules, handicap games, variable Komi options and Byoyomi timed games.
• Hints and undo: request a suggested move or undo instantly, even while the computer is thinking.
• Game records: suspend and restart games, save and load records in SGF files, and replay records automatically or manually.
• Analysis mode: examine the current game or saved SGF records with a move list, histogram and record analysis.
• Other options: automatic territory calculation, highlighting of the last move, a computer resignation feature, and two types of board and stones.