Автор: lawlord
Год: 2026
Языки:
Формат:
Требования:
Ссылки:
Скриншоты:
Описание:
Go 9x9is a board game program that implements Go on a 9x9 board. It features an artificial intelligence that uses Monte Carlo tree search combined with neural network-derived patterns, similar to techniques used by AlphaGo.
The program evaluates positions by simulating approximately 150 games before each move, selecting the move that performed best in these simulations. The AI includes a search tree using RAVE, pattern-based position evaluation trained on a large set of examples, ladder reading, a 300-position opening library, and score estimation at game end. There is a smart time allocation system and area scoring, including rules such as positional superko and komi.
The menu allows for configuration of player colors, move timing, and handicap options. Two-player mode is supported. Version history adds features such as ladder reading and optional joystick input. The program is implemented in assembly for compatibility with specific hardware.