Leela Chess: Installation and Use

In the course of writing my April 2019 review for Chess Life, devoted to the fascinating new Game Changer by Matthew Sadler and Natasha Regan, I had occasion to spend more than a few hours studying Leela Chess Zero, or “Leela.” Like Alpha Zero, Leela is a self-learning algorithm, combining a self-trained neural net and Monte Carlo-style searching to create one of the two or three strongest chess-playing entities on the planet. But unlike Alpha Zero, Leela is open-source and freely available to use by anyone who downloads it. And with its near-win in the TCEC 14 super-tournament, Leela became a source of intense interest among computer chess aficionados and practical players looking for a new analytical tool. This game, in part, stoked that interest.

[pgn] [Event "TCEC Season 14 - Superfinal"] [Site "http://tcec.chessdom.com"] [Date "2019.02.16"] [White "LCZero v20.2-32930"] [Black "Stockfish 190203"] [Result "1-0"] [ECO "C12"] [WhiteElo "3404"] [BlackElo "3588"] [PlyCount "82"] [EventDate "2019.??.??"] 1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. Bg5 Bb4 5. e5 h6 6. Be3 Ne4 7. Qg4 Kf8 8. a3 Bxc3+ 9. bxc3 Nxc3 10. Bd3 b6 11. h4 Nc6 12. h5 Bd7 13. Ne2 Nxe2 14. Qxe2 Ne7 15. a4 a5 16. f4 Nf5 17. Bf2 Nxd4 18. Bxd4 c5 19. Bf2 c4 20. Bg6 fxg6 21. hxg6 Kg8 22. c3 Qf8 23. Qg4 Qf5 24. Qxf5 exf5 25. Rb1 Rb8 26. g4 fxg4 27. e6 Bxa4 28. Kd2 Be8 29. f5 b5 30. Ra1 a4 31. Rhb1 Rc8 32. Bh4 g3 33. Ke3 Bc6 34. Bxg3 Kf8 35. Bd6+ Ke8 36. Rf1 Rd8 37. Bb4 d4+ 38. cxd4 Bd5 39. f6 Bxe6 40. f7+ Kd7 41. f8=R Kc6 1-0 [/pgn]
There are important differences between self-learning engines like Leela and traditional alpha-beta searchers like Stockfish. Some of these differences are easy to grasp, but others require explanation and instruction. Watching this video, the first in a series of occasional efforts for CLO, will tell you exactly what you need to know to install, configure, and interpret Leela’s output.


Links referenced in the video: http://lczero.org/ https://github.com/LeelaChessZero/lc0/wiki/Getting-Started https://github.com/LeelaChessZero/lc0/releases http://lczero.org/networks/ https://github.com/LeelaChessZero/lc0/wiki/Technical-Explanation-of-Leela-Chess-Zero http://forum.computerschach.de/cgi-bin/mwf/topic_show.pl?tid=10194 Game referenced in the video:

[pgn] [Event "URS-ch24"] [Site "Moscow"] [Date "1957.??.??"] [White "Bannik, Anatoly"] [Black "Petrosian, Tigran V"] [Result "1-0"] [ECO "B99"] [PlyCount "51"] [EventDate "1957.??.??"] 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Be7 8. Qf3 Qc7 9. O-O-O Nbd7 10. Qg3 h6 11. Bh4 g5 12. fxg5 Rg8 13. Be2 Ne5 14. g6 Rxg6 15. Qh3 b5 16. a3 Bb7 17. Rhf1 Nxe4 18. Bh5 Nxc3 19. Nxe6 Qc4 20. Bxg6 Ne2+ 21. Kb1 Bc8 22. Ng7+ Kf8 23. Bxe7+ Kxg7 24. Bf5 Nf4 25. Rxf4 Qxf4 26. Bxc8 1-0 [/pgn]

Comments

In reply to by anonymous_stub (not verified)

Good video! One important point that is not clear for me is the selection of the best network to be used at any given time. It's very confusing. ELO evaluation fluctuates a lot, And there are different types of neural networks with different number of layers and nodes. Some of the latest are optimized for slower computers but they are not the strongest ones. Would you please provide advice on this subject?

In reply to by anonymous_stub (not verified)

Network 32930 was used in the TCEC matches and is one of the strongest from series 3. https://lczero.org/get_network?sha=5c222ccd1ccbed2666b3a8ef94d8833d386d2168d51c3e99a0a3fc37a56d2569 The latest 418xx and 419xx nets may be stronger. You can download them at: https://lczero.org/networks/

In reply to by anonymous_stub (not verified)

What are the hardware/software requirements to upload Leela Chess to Fritz 16?

Add new comment

Restricted HTML

  • Allowed HTML tags: <a href hreflang> <em> <strong> <cite> <blockquote cite> <code> <ul type> <ol start type> <li> <dl> <dt> <dd> <h2 id> <h3 id> <h4 id> <h5 id> <h6 id>
  • Lines and paragraphs break automatically.
  • Web page addresses and email addresses turn into links automatically.

Plain Text Comments

Share Your Feedback

We recently completed a website update. If you notice a formatting error on this page, please click here.