Seven Games by Oliver Roeder; a Book Review
A book about computers and pasttimes by the Mathematician-Journalist who started 538's games column.
Created on August 14|Last edited on August 24
Comment
A pastime of pastimes
Tinsley, Carlsen, Sedol, Mochizuki, and Richards are a set of names you may laud as genius incarnate. Alternatively, you may be more taken by Chinook, Deep Blue, Alpha Zero, BG Blitz, Deep Stack, and Maven as their silicone rivals. Regardless of which side of the human-computer battlefield you line up on, you'll be sure to find Seven Games by Oliver Roeder well worth your time.
The book is not poised as a "humans vs. computers" narrative but that undergirds most chapters as a secondary foci. Much of the book instead discusses the culture of competitive play and where the communities stand today. Where vivid, Roeder discusses the histories of the games, and where timely, he engages with the current tournament scene. Roeder spends journalistic effort to meet with and talk to contemporary figures in the games, and he leans on their mathematics background to not shy away from the more analytic components.
Some things I especially enjoyed about this book were the stories about the actual games' masters. Learning more about Tinsley and Mochizuki and others were really enjoyable, and the cultures of these games were also extremely fun to read about. Carlsen’s enthusiasm for Chess AI systems was a surprise to me, and it’s fascinating hearing his mindset of working with them to improve and change his game. I also loved hearing the history of the algorithm development. The book was a bit light on some technical parts that could have been extremely fascinating–things like what changed from one algorithmic approach to the next, discussion of some of the novel data structures in a game like Scrabble–but those sections were excellent when they were discussed.
That said, without a doubt, the biggest let-down in this book was the chapter on Bridge. I'm not a Bridge expert, but the writing in this chapter was dull and lacked nearly all of the respect for the underlying game that the book had for Chess or Go. It was notably the only game of the seven where the author seemed to have not spent much effort themselves trying to learn the game.
For readers of Fully Connected, I have a strong expectation that you'll find plenty of fun things to think about in this book.
Get involved!
In case you're interested, here's a list of open-source AI projects for each of the seven games covered in this book–we look forward to seeing you training some of these yourselves and logging your results to Weights and Biases 😎
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.