Crazy Stone Deep Learning -The First Edition-
Crazy Stone has made a huge step forward by combining Deep Neural Networks with Monte Carlo Tree Search.
Developer
Publisher
Release
May 25, 2016
10 years and 71 days ago
Steam Reviews
Positive35 reviews80.0%
Does this feel indie?
Vote based on whether this game feels indie to you. You can change your vote at any time. Results earn an IGF score after 10 accepted votes.No votes yet
Game contextLatest: IGF Team
| ContextThe evidence factors used to assess whether this game belongs on IndieGame.Fan. | Status | Notes Latest Contributor: IGF TeamEdit notes |
|---|---|---|
Updated: Jul 13, 2026Developer scaleThe operating scale of the game's developer. | ||
Updated: Jul 13, 2026Publisher scaleThe operating scale of the game's publisher. | ||
Updated: Jul 13, 2026Publisher statusHow the game's publishing relationship is classified. | ||
Updated: Jul 13, 2026Creative controlWho appears to retain creative control of the game. | ||
Updated: Jul 13, 2026Release statusWhether and how the game has been released. |
Media
6 itemsAbout the Game
Crazy Stone has made a huge step forward by combining Deep Neural Networks with Monte Carlo Tree Search.
The new Crazy Stone employs Deep Convolutional neural networks as a replacement for the pattern database and has produced a significant improvement in strength.
The new program has scored a winning rate of over 90% against the previous program Crazy Stone 2013 !
We have provided 20 levels of play (13k-7d) for all the board sizes.
Crazy Stone has improved not only in strength, but also in his style of play and the lower levels are perfect for the average players.
Features
- 20 levels of play for each board size (9×9, 13×13, 19×19)
- Human vs Computer, Human vs Human (sharing a single device)
- Computer vs Computer Games
- Handicap games, variable options of Komi
- Hint (suggest)
- Instant Undo (available even when the computer is thinking)
- Automatic territory calculation
- Japanese / Chinese Rule
- Suspend / Re-start games
- Save / Load game record in sgf files
- Automatic and manual replay of a game record
- Highlight the last move
- COM resign feature
- Two types of Board and Stones
- Byoyomi games (Timed games)
(You will not be able to select computer level in timed games) - Analysis Mode (Move list, Histogram, Record analysis)
You can analyze your current game and also game records saved in sgf files. Some screen-shot samples of the analysis mode are shown above.
