Learning from Lc0

For the past few days, I’ve been watching different Youtubers, like GothamChess, agadmator, and chess.com’s Daniel Rensch discuss the gameplay of chess engines like Stockfish and AlphaZero. For those who don’t know, these are some of the top chess engines in the world, with Stockfish 11’s Elo rating being 3558 as of June 2023. In comparison, Magnus Carlsen, the current World Chess Champion, is rated 2830.

As Daniel noticed, Leela Chess Zero’s (another chess engine) (Lc0) moves resemble the Renaissance era of chess, where you would see sacrifices on the board and wild tactical play. What struck me is that the basic principles of the opening, developing your pieces, and taking space, seemed to command the progression of these games. Such play naturally leads you to look for tactics in the middle game. However, the engine also taught me to value piece play so much, that I should even be willing to sacrifice pieces in order to gain superior piece play over my opponent’s. As International Master Daniel Rensch said in his video analyzing the game, this same kind of piece sacrifice theme was found in older master games. Looking out for these themes in my play, and learning how to evaluate the position and determine whether a sacrifice would be worth it is something I want to do. 

I also noticed Stockfish’s and Lc0’s methods of looking at the game were different. These older versions of Stockfish valued material more, while Leela valued positional advantages. This is partly because Leela is a neural network, which is a computer science term detailing that she is made by being fed many, many examples of games, and then determines for herself the type of position she’s looking at. Leela also makes her move based on the probability that she’ll win. I haven’t looked at the complete mechanics behind Leela’s making yet, but her design suggests that Leela has developed a sense of knowing when to execute a piece sacrifice, or initiate aggressive pawn pushes. If we review Leela’s gameplay, I think we can also gain a sense of when and where to make these same decisions.


Posted

in

by