What it teaches
When DeepMind's AlphaZero taught itself chess and crushed the leading engines of the time, its games looked strangely human and startlingly aggressive. Grandmaster Matthew Sadler and Women's International Master Natasha Regan were given access to AlphaZero's games and wrote this study of its style, with support from the DeepMind team.
The book explains how AlphaZero learns, then analyzes its games in depth, highlighting its favorite ideas, such as long-term piece sacrifices, pushing the rook's pawn to attack, and restricting the opponent's pieces, and what human players can learn from them.
Who it's for
Serious players from about 1800 up who want to understand modern, engine-influenced chess, and anyone curious about artificial intelligence. The explanatory chapters are accessible, while the game analysis is advanced.
How to get the most from it
Read the chapters on AlphaZero's key themes, then play through the games with a board, trying to spot those themes before the notes point them out. Look for the same ideas in modern top-level games.
Key takeaways
- AlphaZero values long-term activity over short-term material.
- Pushing the rook's pawn can be a powerful attacking idea.
- Restricting the opponent's pieces is a winning strategy.
- AI has changed how top players think about chess.
Read next
For the story of the first great human-versus-machine match, read Deep Thinking. For modern human master play, try Understanding Chess Move by Move.
