Leela Chess Zero: AlphaZero for the PC

by Albert Silver
4/26/2018 – It might sound like a joke, but it is not: the revolutionary techniques used to create Alpha Zero, the famous AI chess program developed by DeepMind, are now being used to engineer an engine that runs on the PC. This project has now been underway for about two months, and the engine, Leela Chess Zero, is already quite strong, playing at 2700 on good hardware, and is freely available. As will be seen, just about every aspect of this new engine is revolutionary and breaks all the paradigms that had dominated the field until now.

ChessBase 14 Download ChessBase 14 Download

Everyone uses ChessBase, from the World Champion to the amateur next door. Start your personal success story with ChessBase 14 and enjoy your chess even more!

Along with the ChessBase 14 program you can access the Live Database of 8 million games, and receive three months of free ChesssBase Account Premium membership and all of our online apps! Have a look today!


From Alpha Zero to Leela Zero

When Alpha Go, and then Alpha Go Zero, was released, it cracked one of the most challenging strategic games in the world, and one that experts all over had thought would resist for at least another decade. It was the result of a powerful artificial intelligence technique known as deep reinforcement learning, which built a neural network combined with the Monte Carlo Tree Search that allowed the program, equipped with just the barebones rules of the game, to play itself millions upon millions of times, learning as it went until it reached, as was the case, superhuman playing strength.

All the more remarkable was that the Deep Mind group published peer-reviewed papers that detailed exactly how it was achieved. In a sense, this was the equivalent of IBM creating the world-beating Deep Blue, and then publishing blueprints the very same day so that anyone could build their own. As a sidenote, Feng Hsu and the Deep Blue team did eventually publish papers describing the entire thing inside and out. Nevertheless, it was believed that while a game such as Go, might fit this form of software engineering, it would never really work for a chess program. At least not at the highest level. After all, Go is less about calculating exact lines, and more about extensive pattern recognition, while chess is highly tactical and seeing just a single extra move ahead can make or break a program.

Then along came Deep Mind, who poured in their resources and built a world-beating engine that learned the game itself within 24 hours. Naturally, there were plenty of special conditions in play, such as having access to thousands of incredibly powerful processors to help develop its ability in such a short time, but they did it. Consider that if Murray Campbell and Co. had thought that just piling in special hardware would somehow solve chess for them in 24 hours, they would have done it. Deep Blue had a system performance of 12 GFLOPS, while the horsepower DeepMind used easily exceeded 40 million GLOPS.

DeepMind's professor David Silver explains the new 'Zero' approach in AlphaGo Zero, which preceded Alpha Zero (chess)

The new Alpha Zero chess program lead to an astounding media frenzy, and just as much controversy in the chess world. Much was made about the conditions of the match against a 64-thread version of Stockfish used to test its strength, but this was to completely overlook the important aspect of it all: a new world-class engine, using a completely new method and technique, had been developed in 24 hours. What was more, anyone was invited to make their own.

The paper "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm" at Cornell University

This was not nearly as easy as might seem. For one thing, a single person could hardly do it at home with a single machine, or even a room full. Then there is the matter of trying to translate Deep Mind’s explanations in their paper into concrete computer code.

But some programmers took up the challenge, and after developing a Go program, Leela Zero, the strongest one publicly available now, based on community effort and contributions, it has been transferred to chess.

Leela Chess Zero is based on the seminal Alpha Zero paper, using the same self-learning techniques and the same Monte Carlo Tree Search technique. It is hosted at the LCZero site and there one can not only find the engine but the latest version of the neural networks — the brain if you will.

How strong is Leela Chess Zero?

The project is still in its infancy, having only really started two months ago. As it depends very much on help by others to contribute computer time, the beginning was slow as time was needed for word to get around, tools developed to make it easy, etc. Because of the unique nature of the engine’s design, using a huge evaluation function (over 80MB), it gains exponentially in strength when used with a good graphics card.

Without a graphics card, using only my CPU, I barely get 20% the performance I get with the GPU (graphics processing unit). On a laptop equipped with a quad-core i7 and a fast GTX980M graphics card, it plays at around 2700 Elo or so. Remember that this is after just two months of self-learning. However, the engine’s play is quite different from anything I have seen or tested before, and I am speaking from over 20 years experience.

Without exception, until now, an engine’s strength was clearly determined by its speed and tactics. Sure an engine could compensate a slight tactical deficiency with more knowledge, but not by much, and more often than not, the ‘dumber but faster’ engine would compensate its ignorance with slightly deeper calculations. This give and take has defined engines on the PC since the beginning of computer chess.

Therefore you can imagine what I thought after the following: I gave a fairly weak engine — Firefly (rated 2250) — a basic tactics test of 100 positions, a set so easy I would expect myself to score 100, and it scored 94/100. I then fed it to Leela Zero, which barely scored 41/100. I grimaced, knowing this meant Leela was going to get trounced, but still decided to play a small match of 10-minute games just to gauge its play. I found myself forced to stop the match early after it was 10-0 for…. Leela!?! I check and triple-checked the settings to make sure everything was fine. 

I then tried another engine, Delphil 3.2 (rated 2494) with the same 10-minute conditions, and stopped it after 20 games with Leela bullying it 15½-4½. What on earth was going on?? I never heard of an engine compensating hundreds of Elo of tactical gaps with pure positional play. That is a human thing, not a machine. Eventually, it found its match with Nebula, a 2700+ engine with a  14½-14½ score, enough to suggest this is about where it stands.

With a description of an engine compensating tactical weaknesses with massive positional play, you might expect games that looked like Carlsen or Capablanca, with controlled complications, and then a boa constrictor win later, right? Not even close! Leela Zero makes Topalov, the king of exchange sacrifices, look like a timid milkmaid. It plays a ton of speculative moves, readily giving up pawns, exchanges and more for activity and play. Please note the word speculative, because I doubt very much these are all correct. However, they do make for very entertaining games, and what is more, it plays the endgame impressively well.

Take a look at the following game, which sort of illustrates some of the surprises it can come up with:

Nebula 2.0 vs Leela Chess Zero (NN189)

[Event "Match"] [Date "2018.04.25"] [Round "6"] [White "Nebula 2.0 64bit"] [Black "Lczero 0.7 NN189"] [Result "1/2-1/2"] [ECO "A30"] [WhiteElo "2726"] [BlackElo "2700"] [PlyCount "191"] [EventDate "2018.??.??"] 1. c4 c5 2. Nf3 Nf6 3. g3 b6 4. Bg2 Bb7 5. Nc3 e6 6. O-O Be7 7. d4 cxd4 8. Qxd4 d6 9. Rd1 a6 10. b3 Nc6 11. Qd3 Qc7 12. Bf4 Rd8 13. Qc2 O-O 14. a3 Nb8 15. e4 h6 16. Rac1 Nbd7 17. h3 Nc5 18. Re1 Ncd7 19. Be3 Rfe8 20. Nd4 Qb8 21. f4 Rc8 22. Rcd1 b5 $5 {[#] It is more than a little astonishing to see an engine play this thematic double-strike with ...b5 and ...d5 in the hedgehog. It isn't quite correct, though Black does get play for it.} 23. cxb5 d5 $1 {In for a penny, in for a pound.} 24. e5 Ne4 25. Bxe4 dxe4 26. Nc6 Bxc6 27. bxc6 Rxc6 28. Rxd7 Bxa3 29. Rd4 Rec8 30. Rc4 Rxc4 31. bxc4 Rxc4 32. Qb1 Rb4 33. Qa2 Bb2 34. Nd1 Qb5 35. Kg2 h5 36. Kh2 Rb3 37. Bf2 Rb4 38. Kg2 a5 39. Nxb2 Rxb2 40. Qa3 Rb3 41. Qe7 Qc6 42. Qc5 Qb7 43. Qa7 Qc6 44. Be3 a4 45. g4 h4 46. Rc1 Qd5 47. Qe7 Rb8 48. Qxh4 a3 49. Qe1 Qd3 50. Rd1 Qb3 51. Qd2 a2 52. Ra1 Ra8 53. Qc1 Rd8 54. g5 Rd3 55. Bf2 e3 56. Be1 Qb7+ 57. Kh2 e2 58. Rxa2 Qf3 59. Qc8+ Kh7 60. Qc2 Qxh3+ 61. Kg1 Qf1+ 62. Kh2 Qh3+ 63. Kg1 Qg4+ 64. Kh1 Qf3+ 65. Kg1 Qf1+ 66. Kh2 Qxf4+ 67. Kg2 Qxg5+ 68. Kh1 Qh6+ 69. Kg1 Qe3+ 70. Kg2 Qe4+ 71. Kf2 Qf4+ 72. Kg2 Qf3+ 73. Kg1 Qg4+ 74. Kh2 Qh5+ 75. Kg1 Qg6+ 76. Kh2 Qh5+ 77. Kg1 Qg5+ 78. Kh1 Qh6+ 79. Kg1 Qe3+ 80. Kg2 Qe4+ 81. Kf2 Qd5 82. Ra4 Qf3+ 83. Kg1 g5 84. Ra3 Qf1+ 85. Kh2 Qxe1 86. Qxd3+ Kh6 $4 {An unfortunate blunder that allows White to save the game.} ({After} 86... Kg7 $1 {Black would win.}) 87. Qd4 Qf1 88. Rg3 e1=Q 89. Qh4+ gxh4 90. Rg6+ Kh7 91. Rg7+ Kh8 92. Rg8+ Kh7 93. Rg7+ Kh6 94. Rg6+ Kh5 95. Rg5+ Kh6 96. Rg6+ 1/2-1/2

Bear in mind that while it misses some easy tactics it also catches some hard ones very quickly.


In the position above, White (Leela) played the lovely 38.c5!! And went on to win after 38…dxc5 39.Rd7!

Interview with Gary Linscott

I caught up with Gary Linscott (pictured), the person who spearheaded the Leela Chess Zero project and helped make it what it is today. (Please note that another interview with Gian-Carlo Pascutto, creator of Leela Zero — which led to Leela Chess Zero — will appear soon.)

Albert Silver: Hi Gary, congratulations on the fascinating project. How did it come into existence? Concept, execution, etc

Gary Linscott

Gary Linscott: The AlphaZero paper was a landmark — it was the first time traditional techniques had been defeated in over 20 years. I had wanted to learn about reinforcement learning, and this provided a great motivation. I even started a small version of this project in python with a co-worker, but we weren't generating anywhere near enough games to make reasonable progress. Then, Leela Zero really started to take off. A user named Benediamond got the ball rolling with a non-working prototype of chess support for Leela Zero. It needed a fair bit of additional work, which I did, and then coded up the server and client in Go [the coding language, not the game -Ed.] to learn more about that language. Then, I launched it on Talkchess, and the support has been incredible since then!

AS: How is Leela Chess Zero being developed? Is it just you, or others, or...? 

GL: There is an extremely active community! Folkert Huizinga (Error323) was one of the earliest to join and has been the driving force behind the training process, which generates the new networks. It's an extremely hard, and time-consuming part of the project, and he's done an amazing job. We got a donation of the training hardware as well, which has been fantastic. Then a bunch of great developers have stepped up to help out as well, special mention to Andy Olsen (KillerDucky) and Alexander Lyashuk (mooskagh).

AS: How close is it to Alpha Zero, or how different? 

GL: It's quite close to AlphaZero in concept. The paper doesn't really reveal all the little details, so we aren't 100% sure. For example, there is a critical change in the network architecture that was required for chess that wasn't in the paper (the number of layers in the policy head for the technically inclined).

AS: What has been the biggest challenge so far in creating/developing it?

GL: Keeping up with the community! There is a lot of passion, but it's a part-time project on my side. Others have really stepped up and made huge contributions thankfully.

AS: Roughly how strong is Leela Chess Zero right now in FIDE Elo terms?

GL: It depends (smiley). It's very sensitive to the hardware, and the time control used. However, in recent blitz/bullet matches, it was able to defeat human IM/GMs convincingly. Against other machines, it's still a long ways from the top — estimates vary between 2600 and 2800, while something like Stockfish is ~3300.

AS: How much strength does a decent video card add? 

GL: A huge amount! For example, the TCEC competition is playing it on ~40 cores, and it would be roughly the same speed on a 1080Ti.  We will also improve the efficiency significantly on GPU as time goes on, with things like CUDNN, and TensorRT.

AS: What has been the biggest surprise so far?

GL: How much interest and passion there is around this project. It's been fantastic to watch the community really get excited, and all the chess personalities who have tried playing it.

AS: While testing the engine in Fritz, I noticed a number of very strange tactical oversights, even in fairly simple positions. Is this normal, and if so, when would they be overcome?

GL: Definitely normal, the network is still learning tactics, especially at low depths. Also, we are on a much smaller network than DeepMind used (10x128 instead of 20x256).

AS: How many games did AlphaZero need to be trained to its famous level, and how many has Leela Chess Zero used? Also, at the current rate of development, how long would it take to reach that level (presuming similar hardware)?

GL: From their paper, 44 million if I recall correctly, while we are at 8.2 million as of this email. However, we've also had some significant bugs during the training process, and have been learning a lot as we go. I wouldn't be surprised if our run took longer.

AS: Can you tell a bit about yourself

GL: I've always loved the combination of distributed systems and chess. A few years back I also developed "fishtest", which helped take Stockfish to number one on the rating lists (although these days, a wonderful community has taken it on). My professional work is on developing autonomous cars, which is extremely fun and challenging.

How to install Leela Chess Zero in Fritz/ChessBase

Leela Chess Zero can be installed as a UCI engine in Fritz or ChessBase, though it does require a few extra steps. Also, you most definitely want a video card to run it properly. The faster the better, but anything beyond Intel integrated graphics should be a plus.

1) Download the latest Windows build (there are Mac versions too), it comes with the needed DLLs: https://github.com/glinscott/leela-chess/releases.

If you have a discrete GPU (graphics card), get the gpu-win.zip version instead

Unzip into a folder of your choice. I created one called \LCZero.

2) You may also need to download and install the VC Redistributable 2015 x64. Many other programs rely on this library so you may have it already.

3) You will now need to double-click on the LCZero.exe engine one time. It is vital this be done before trying to install it as a UCI engine.

4) Go to the Network page and download the latest network file. Naturally, you can download any version really, even when it was rated a big fat zero. Right-click on the blue Network number and save.

You then unzip it to the main Leela folder. Now you will have a very large file with a long name that looks like a secret code. Rename it to weights.txt

5) If you have a GPU, then I strongly suggest you download this .bat file and unzip it in the same folder as Leela and then run it. This is a simple command that will have the computer run a full tuning of your computer to get best results with Leela. It may take 20-30 minutes, so be patient.

Running the script will do a full tune of your graphics card to get optimal results. Please note this can improve performance to almost double even!

You are now ready to install it in Fritz/ChessBase!

6) In Fritz/ChessBase, select Create UCI engine 

Then find LCZero.exe file, and that is it!

When you want to update the neural network with the latest version, all you need to do is download the newest Network file, and rename it again.

Contributing to Leela Chess Zero

If you would like to contribute some of your computer time to help Leela, just double-click on the client.exe file in the folder. It will start running self-play games, and send them to the site. To stop the process, just close the window. For more information, please see the Getting Started page.

Questions and suggestions

If you are having trouble with Leela, or have any questions or suggestions, be sure to check out the official forum for it.


Born in the US, he grew up in Paris, France, where he completed his Baccalaureat, and after college moved to Rio de Janeiro, Brazil. He had a peak rating of 2240 FIDE, and was a key designer of Chess Assistant 6. In 2010 he joined the ChessBase family as an editor and writer at ChessBase News. He is also a passionate photographer with work appearing in numerous publications.
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Rules for reader comments


Not registered yet? Register

RJurjevic RJurjevic 5/12/2018 10:51
Test game 4. Vafra 2.11.0 vs LeelaZero 0.8, a proper slow time control chess game, on HP Pavilion dv7 laptop, using Intel® Core™ i7, with four hyper-threaded Cores, no GPU used, probably because there isn't one, which did not suit LeelaZero, I understand. Both engines thought only on their own move, i.e., no pondering, used six threads, the same opening book, Fritz 13, with default settings. Vafra had access to endgame tablebases, all five- and some six-men. Leela was using a brain of 12/May/2018, i.e., NN281.

RJurjevic RJurjevic 5/12/2018 09:36
Test game 3. Vafra 2.11.0 vs LeelaZero 0.8, a proper slow time control chess game, on HP Pavilion dv7 laptop, using Intel® Core™ i7, with four hyper-threaded Cores, no GPU used, probably because there isn't one, which did not suit LeelaZero, I understand. Both engines thought only on their own move, i.e., no pondering, used six threads, the same opening book, Fritz 13, with default settings. Vafra had access to endgame tablebases, all five- and some six-men. Leela was using a brain of 09/May/2018, i.e., NN278.

RJurjevic RJurjevic 5/11/2018 10:43
@Alpha Zero and others: AlphaZero, as far as I understand, uses deep neural network learning, and a variant of, I think, UCT or the like, Monte Carlo tree search. The UCT, or the like, part is here, as far as I understand, to avoid randomness in the search. Let us assume that in a position a move performs very well on average, but with precise play can be easily refuted, and that other move can be proven to be better. UTC, or the like, I think, addresses this problem, i.e., allows Monte Carlo not to play purely randomly, but with some 'knowledge' of which moves are more promising. In AlphaZero, I think, they also use deep neural network 'knowledge' to aid to choosing of 'good' moves. Now, I am not sure if AlphaZero, when it plays, chooses the moves based on the neural network knowledge only, or also on the UCT, or the like, Monte Carlo real time tree search as well. In a nutshell, as far as I understand, artificial neural networks resemble real neural networks, say in our brains, and have myriad of weights, i.e., numbers, to be set. These weights define how the network will process the input, i.e., what will be the output for a given input. I think that in AlphaZero, eventually, the network is used to predict the best move for a given position. So, in order to train the neural network, one needs to have sample of data on which to check how the network performs, i.e., what it knows, and to adjust the weights to fit the data as good as possible. It would appear that the data set they use for AlphaZero, is UCT, or the like, Monte Carlo tree search in self play. Then, based on the data, the neural network weights are adjusted, or I think they say fed back. Then a new set of data is generated, using self play, but this time with a more knowledgeable neural network, i.e., on average better moves are played, and the feed back, based on this new data, further increases the knowledge of the network, etc., and the process is repeated. For Leela one can download ``the latest version of the neural networks — the brain if you will''. What they meant by that is, I think, that one can download a set of neural network weights, which are obtained by self play. That is, I think, why they need volunteers to give their CPU or GPU time, i.e., to generate the data based on self play, which is then used to tune the network knowledge. That is why AlphaZero needs no other knowledge, say such as that Rooks are good on open files, etc., than the rules of the games, which include which moves are legal in a position, what is checkmate, stalemate, etc.

RJurjevic RJurjevic 5/11/2018 09:44
@crispyambulance: GPU can be programmed to calculate other things, not only for graphics, etc., and are, as far as I understand, good for simple massively parallel tasks.
RJurjevic RJurjevic 5/11/2018 08:18
Test game 2. Vafra 2.11.0 vs LeelaZero 0.8, a proper slow time control chess game, on HP Pavilion dv7 laptop, using Intel® Core™ i7, with four hyper-threaded Cores, no GPU used, probably because there isn't one, which did not suit LeelaZero, I understand. Both engines thought only on their own move, i.e., no pondering, used six threads, the same opening book, Fritz 13, with default settings. Vafra had access to endgame tablebases, all five- and some six-men. Leela was using a brain of 09/May/2018.

RJurjevic RJurjevic 5/11/2018 03:05
This may be of some interest.

RJurjevic RJurjevic 5/11/2018 02:02
Test game 1. Vafra 2.11.0 vs LeelaZero 0.8, a proper slow time control chess game, on HP Pavilion dv7 laptop, using Intel® Core™ i7, with four hyper-threaded Cores, no GPU used, probably because there isn't one, which did not suit LeelaZero, I understand. Both engines thought only on their own move, i.e., no pondering, used six threads, the same opening book, Fritz 13, with default settings. Vafra had access to endgame tablebases, all five- and some six-men. Leela was using a brain of 09/May/2018.

RJurjevic RJurjevic 5/11/2018 01:57
Thank you. Great stuff!
celeje celeje 5/6/2018 12:50
@puertolajas: Sounds like easiest is to use a different version of Fritz.
Yeah, I know, not helpful if 16 is the only one you have or want to use.
puertolajas puertolajas 5/5/2018 11:18
Is there anyone who has been finally able to run Leela Zero under Fritz 16?? It seems like mission impossible!! Not even following the instructions on Chessbase.com
celeje celeje 5/1/2018 04:03
The NN size was upgraded again - 2nd time. So it's more time-consuming again.
Looks like games contributed number is about 2/3 what it was before.
ryuken ryuken 5/1/2018 06:42
@ Albert Silver: ive installed leela on fritz and it runs fine. my only problem is that it is showing 0 knps when analyzing and playing games. and there is no display on moves being analyzed currently. why is that? i just followed your instructions above.
garyklien garyklien 4/30/2018 06:51
Not sure as I only got it to work for a few hours one day, about 140 games. Then the next day it constantly tells me it cant find leelachess.exe file. So....Running someone my rig atm, but cant do nearly as much as if the coalab worked for me and didn't disconnect after 20 games when it does work.
celeje celeje 4/30/2018 05:56
@ garyklien: the fundraising page says they want to test its performance before buying. That's just for training.

Creating self-play games needs the community's computer time.
Reported numbers are confusing.
Anyone can use the free Google Colab Tesla K80 GPU
The lc0 page says this creates 1000-1500 games a day.
But some say their laptops create 1000 games a day.
Dunno how both numbers can be right.

How many self-play games is your computer playing?
garyklien garyklien 4/29/2018 04:06
Cele I thought the supercomputers for chest did better when they bought mini lower-powered processors are gpus to run together
celeje celeje 4/29/2018 12:06
The lc0 fundraising campaign (gofundme.com/leela-chess-zero) for training hardware to run 24/7 is also going well. They're considering e.g. Titan V.

Hope everyone is patient & doesn't push for addons to current self-training. That'd just dilute available hardware.
fons3 fons3 4/28/2018 10:42
Johannes Kiliani on why LC0 does badly on tactics puzzles:

Leela will always be weak(er) playing positions that are very unlike those encountered in the actual training games, since here the neural net has to make deductions from insufficient good information. The policy head only has to miss one important candidate move in a tactical line for the engine not to find that tactic. In regular games against humans or other engines, it's very difficult for the opponent to lead Leela far enough out of her depth for this effect to matter without getting into a losing position anyway, but for a chess problem that she never had any contribution getting into? Likely happens all the time that those just don't look like anything Leela was ever trained on, with predictable results.

At some point in the future, it could probably be tried to train Leela (or a fork of her) on a mix of self-play games and tactical suites like this one. There's a chance that this would make the net at least more familiar with these types of positions, which is what's required to solve such puzzles. May take quite a bit of trial and error though to do that without weakening the general gameplay of the neural net, and it will definitely require a very LARGE database of tactical puzzles to train on.

(from http://talkchess.com/forum/viewtopic.php?start=0&t=67239&topic_view=flat)
garyklien garyklien 4/28/2018 07:36
Hi, Bertman. I belive this should be added as it will be the main way most can connect, otherwise it would just be chessbase users.


Shows how to let google colab processors run for you, instead of running your personal computer, or you can do both!
mpfau mpfau 4/28/2018 06:13
Fascinating article, very enjoyable read.
HouseMaster HouseMaster 4/28/2018 04:54
although Leela works perfect in other GUI, not in Chessbase! Can someone make a prepared folder, with a working Leela? Thank you!
puertolajas puertolajas 4/28/2018 04:08
The Network page is not working! Does anyone have the same problem?
feldi feldi 4/28/2018 11:23
What if we combine Leela Chess Zero with a tactical strong "traditional" engine, like stockfish? E.g. like a "Zweihirn"? Wouldnt that be the strogest engine ever? Any plans to do that?
vikas2200 vikas2200 4/28/2018 08:55
Can I install Leela Zero on 32bit windows OS as well.
celeje celeje 4/27/2018 11:04
@ RayLopez:

As Bertman said, no alpha-beta.

Calling a sacrifice "positional" is just the human way of thinking about it. A computer program does not "think" "positional" vs "tactical". They are all the same thing for the program.
garyklien garyklien 4/27/2018 09:54
I wish we could all at ChessBase login to one name and then see how many games we contribute as a whole :) like the google group coop :)
Bertman Bertman 4/27/2018 09:06
No, Leela does not use a classical tree search at all, and explicitly does not use an alpha-beta search. It uses a general purpose Monte Carlo Tree Search.
flachspieler flachspieler 4/27/2018 08:16
Mama Mia, what a development.
RayLopez RayLopez 4/27/2018 08:07
@celeje - Re speculative, what you say doesn't contradict anything I said. A positional sacrifice by definition is 'speculative'. BTW AlphaZero-inspired Leela Zero is in fact using 'alpha-beta' for all plies. The difference with traditional programs is just in the move ordering algorithm and how much weight is assigned to each branch of the chess tree. Instead of just 'static' rules like "centralize the knight", in addition, Leela uses a knowledge database based on actual games played (and some sort of hash table of positions I'm pretty sure) so that in a position that cries out for a Greek Gift (speculative) sacrifice, that does not lead to immediate mate but puts permanent pressure on the opponent, the Leela program will make such a sacrifice, whereas a traditional program may hesitate since there's no immediate gain in centipawns.
garyklien garyklien 4/27/2018 05:29
Got it working, adding to this leelachess thing. Thanks for giving info on how to do this :) If we get enough readers to do a lil bit at a time could see some nice improvements fairly soonish.
KWRegan KWRegan 4/27/2018 03:13
The global "ArXiv" service is maintained by Cornell University but there is no other association between the AlphaZero researchers and Cornell. I am delighted that the Leela Zero project has reached this threshold.
melomaniak melomaniak 4/27/2018 02:14
Can it use multi-GPU?
celeje celeje 4/27/2018 01:40
@ RayLopez:

It's only speculative for humans. The program does not 'think' it's speculative. The program can be wrong, but it's not deliberately being speculative. If it calculates much further & better, it just looks speculative.

Junior is an alpha-beta program and did 'speculative' sacrifices more than a decade ago.
RayLopez RayLopez 4/27/2018 12:59
@celeje - I beg to differ. Though I've tried my hand at computer chess programming, I'm not an expert, however, I do know enough that the standard "Alpha-Beta" algorithm, with pruning, generally does not, for move ordering, employ speculative sacrifices, except (from what I've seen) maybe a piece for two pawns. By contrast, the Leela AlphaZero variant is using Monte-Carlo to uncover deep positional sacrifices beyond the move-ordering rules used now. Today's rules of "centralize knight", "rooks behind pawns", "king attacks in endgame with queens off", etc, is not enough, and Leela is showing that's true. So, Tal was right, there's more to chess than the standard numerical values of the pieces and the traditional rules I outlined above.
celeje celeje 4/27/2018 11:41
@ RayLopez:

It's "Tal-type" if a human plays it, but not Tal-type if a computer plays it.

On your second point: there was nothing really new in the AZ project. It was just an application of known NN stuff. The real issue is getting enough computer time (so everyone interested should donate computer time to lc0).
RayLopez RayLopez 4/27/2018 11:17
What's interesting to me is how Tal-type sacrifices beat the PC, something that you figure out whenever you play vs a PC and make a sacrifice just outside the 'horizon effect' of a chess program. This works even for computers playing computers. This shows the chess tree is conducive to these type sacrifices, and that "Tal was right"! A second interesting point is how these open-source programmers managed to recreate AlphaZero from scratch, since the published paper was sparse on details, except for playing many games from a node on the chess tree and storing the results.
Pacheco76 Pacheco76 4/27/2018 07:40
Since you mentioned Stockfish: Any plans to cover the TCEC final? Or does it only happen when Komodo or Houdini win?
Alpha Zero Alpha Zero 4/27/2018 07:01
As I understand it, the whole idea behind Alpha Zero is that it improves by playing millions of games. So if you beat Leela Chess in the evening and you just leave her, turned on, over night, she'll beat you big time in the morning?
garyklien garyklien 4/27/2018 05:41
I am excited to add games to the database. Running the testing.bat program as wel speaks
Bertman Bertman 4/27/2018 04:55
@TMMM - You also need to take into account that DeepMind was running it on hardware that ran at 180 TFLOPs. As a comparison, a GTX 1080ti, currently retailing for $1000 can do about 11 TFLOPS.

Anyhow, on the same machine, I let Komodo 11.2 analyze it, and it took over a minute to find 38. c5!!, so it is a challenging move for engines in general, and shows that while Leela may miss easy shots still, it is also capable of finding moves others miss in similar time.

I will add that I think the slightly imbalanced and very dynamic style must make it a fun opponent to play against, the same of which cannot necessarily be said of all engines, regardless of their strength. Just my 2 cents.
LetoAtreides82 LetoAtreides82 4/27/2018 03:01
There's a gofundme page for this project, see here: