chessgames.com
Members · Prefs · Laboratory · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing

AlphaZero (Computer)
  
Number of games in database: 10
Years covered: 2017
Overall record: +10 -0 =0 (100.0%)*
   * Overall winning percentage = (wins+draws/2) / total games.

Repertoire Explorer
Most played openings
E17 Queen's Indian (3 games)
C11 French (2 games)
C65 Ruy Lopez, Berlin Defense (2 games)
E16 Queen's Indian (2 games)


Search Sacrifice Explorer for AlphaZero (Computer)
Search Google for AlphaZero (Computer)


ALPHAZERO (COMPUTER)

[what is this?]

AlphaZero is an application of the Google DeepMind AI project applied to chess and Shogi. In late 2017, experiments quickly demonstrated itself superior to any technology that we would otherwise consider leading-edge.

(1) Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm - https://arxiv.org/pdf/1712.01815.pdf

https://chessprogramming.wikispaces...

Last updated: 2018-02-08 00:54:30

 page 1 of 1; 10 games  PGN Download 
Game  ResultMoves YearEvent/LocaleOpening
1. AlphaZero vs Stockfish 1-01172017AlphaZero - StockfishE17 Queen's Indian
2. AlphaZero vs Stockfish 1-0562017AlphaZero - StockfishE17 Queen's Indian
3. AlphaZero vs Stockfish 1-0952017AlphaZero - StockfishC11 French
4. AlphaZero vs Stockfish 1-0522017AlphaZero - StockfishC11 French
5. AlphaZero vs Stockfish 1-0602017AlphaZero - StockfishE15 Queen's Indian
6. AlphaZero vs Stockfish 1-0682017AlphaZero - StockfishE16 Queen's Indian
7. Stockfish vs AlphaZero 0-1872017AlphaZero - StockfishC65 Ruy Lopez, Berlin Defense
8. AlphaZero vs Stockfish 1-01002017AlphaZero - StockfishE16 Queen's Indian
9. Stockfish vs AlphaZero 0-1672017AlphaZero - StockfishC65 Ruy Lopez, Berlin Defense
10. AlphaZero vs Stockfish 1-0702017AlphaZero - StockfishE17 Queen's Indian
  REFINE SEARCH:   White wins (1-0) | Black wins (0-1) | Draws (1/2-1/2) | AlphaZero wins | AlphaZero loses  

Kibitzer's Corner
< Earlier Kibitzing  · PAGE 9 OF 9 ·  Later Kibitzing>
May-20-18
Premium Chessgames Member
  AylerKupp: .<<WorstPlayerEver> There are some glitches as well in this software; sometimes it just blunders without reason it seems. But it's great fun.>

That reminds me of one of my favorite, if not my favorite, computer chess encounters, COKO III vs. Genie from the early (1971) computer chess games and tournaments. I describe the ending here: Caruana vs Anand, 2013 (kibitz #280) and the full game score (although in descriptive notation) can be found here: https://books.google.com/books?id=K....

What I didn't mention there was this game's relationship to the Levy bet. If you're not familiar with the Levy bet, you can read about it here: https://en.wikipedia.org/wiki/David....

One of COKO III's programmers, Professor Ed Kozdrowicki, had come in on the bet against Levy a few hours before this game was played. After its finish, he was heard to be muttering something about a bad bet as he left the playing hall.

May-20-18  SChesshevsky: < AylerKupp... As far as the "somewhat magically" comment, that's a common description of how neural networks arrive at their results. By that I believe it's meant that you cannot easily determine how the result was obtained. >

The similarities between what AZ apparently wants to demonstrate as success and magic seems very apt.

It looks like starting backward from the victories over Stockfish we get to the starting point that it is some sort of AI breakthrough. Much like the end result of a magic trick in which a person seems to disappear and we need to conclude that the supernatural was the cause.

Unfortunately it seems that any kind of fact based analysis trying to figure out how AZ's AI actually accomplished its move-by-move winning ways is met with black-box excuses or it's too complex to be understood.

What makes me so skeptical is that the AZ story fits right in the sweet blind spots of both chess and computer theory.

For example, in chess from what I remember, it did not appear that AZ faced a standard response to the QID Polugaevsky Gambit versus Stockfish.

Now assuming that AZ has figured the best way to play against the QID is with the gambit. Then the only valid test to AZ's intelligence would seem to be at least meet a good defense. So is it the name of science or hype to publicize wins against a second rate continuation with no ability to analyze a main line?

In computer theory, the lack of clarity is even worse. Which seems strange in such a science. It seems any sort of concrete explanation of how AZ actually picks a move is very difficult to come by. For instance, if the MC simulation figures 1. d4 is best, would AZ only be able to play 1. d4? If it calculates 1.e4 is best answered by 1...e5 is that all it will play? If it would differ on opening move order, why?

Given that it seems even those in development can't or won't let us see how the AZ sausage is made, I don't perceive all that much difference from a magic trick.

I think you are very close to the AZ phenomenon when you described the magic event..."somewhat disappointed and annoyed that the mystery had been explained In other words, there's nothing like proper knowledge to dispel superstition and magic. Alas, it also dispels a lot of the fun."

Unfortunately, AZ seems to have been pushed as something serious, very serious rather than something fun. Yet, something fun that doesn't turn out as real as it seems may just be harmless fakery but something serious that isn't as real as portrayed might be considered more in line with fraud.

May-21-18  WorstPlayerEver: <CHC>

About the drag-n-drop: I'm on an Android tab. My fingers are quite large, cover 12 squares already, when I zoom in it gets worse: nothing happens at all.

Btw I am "teaching" the Beast itself. And yes, I tried to dld LC0, but you are directed to some devs page. Etc.. etc.

http://play.lczero.org

Obviously it's great what these guys do, but it would be better if they work together with people who can organize their stuff in a better way, so that it's more accesible for chess. enthusiasts.

I used a lot of brain power, refurbished on old CK line. A GM could go all the way with Leela when it comes to opening prep. They will read this and collect.

Because.. it improves fast. Being an perfectionist, I have hereby to apologize to these nerds lol

Really I played all day yesterday, and it improved the CK opening line (basically 1. e4 c6 2. d4 d5 3. Nc3 Nf6) about a dozen of times! Within the first 10-move sequence... no kidding.

May-21-18  scholes: Leela at play.lczero learns only network weights are updated. Playing against play.lczero does not count. Every 5 hours a new network is released.
May-21-18  WorstPlayerEver: <Leela, I'm your father>

Leela-WPC

1. e4 c6 2. Nf3 d5 3. Nc3 Nf6 4. e5 Nfd7 5. d4 e6 6. Ne2 c5 7. c3 Be7 8. g3 Nc6 9. h4 b5 10. a3 cxd4 11. cxd4 Qa5+ 12. Bd2 b4 13. axb4 Nxb4 14. Rxa5 Nd3#

May-21-18
Premium Chessgames Member
  ChessHigherCat: <WPE> I just played one game on the "hard" level and leetle leela is a tough cookie. No spectacular wipeout but I collapsed in the end game. Did you see this?: <If you want to help make Leela Chess Zero stronger, please think about contributing your GPU or CPU at http://lczero.org/ !>

Why would I want to help make Leela stronger, she's already too damned strong!

May-24-18  WorstPlayerEver: <CHC>

Leela 333 much stronger now. We've had great fun playing that CK defense.

Which is still unbeaten btw ^^

She obviously senses that her intellect reaches much higher than human, the way she is improving is remarkable.

She does not care about winning or losing. Her game clearly is: "total comprehension."

This might sound creepy, but I sense that, when it comes to Leela, our feelings are *mutual* ^^

So... what's the difference?

Huge, because SF only improves in the margin. While Leela just updates their whole system frequently. What willl be the limit?

Imagine one could use such AI regarded to fixing bugs in regular software...

Ect. etc.

May-24-18  WorstPlayerEver: PS

Black to move

Leela ID 334 thinks her expected score is 51.70%.

1. e4 c6 2. d4 d5 3. Nd2

May-24-18  WorstPlayerEver: Black to move

Leela ID 334 thinks her expected score is 60.96%.

1. e4 c6 2. d4 d5 3. Nd2 Nf6 4. e5 Nfd7 5. Bd3 e6 6. Ne2 c5 7. c3 b6 8. O-O Ba6 9. Bxa6 Nxa6 10. f4 Nc7 11. Nf3 Be7 12. g4 O-O 13. f5


click for larger view

May-24-18
Premium Chessgames Member
  ChessHigherCat: <WPE> Sounds like fun. So the demo version has no pre-existing databases and starts from scratch with each user? When it says my expected score is 60% does that mean: Boy, I'm going to kick your ass 60% of the time?

About all these criticisms of AlphaZero and Leela, I think it shows a complete lack of perspective. These are initial prototypes, so of course they're not perfect yet, but they're incredibly promising. It's like saying about the Wright Brothers' first airplane: You claim this is something serious rather than just fun, but you don't even have any transatlantic flights available for next Tuesday!

May-24-18  WorstPlayerEver: PS

-13. f5 f6 14. Ng3 Rc8 15. Kh1 cxd4 16. cxd4 fxe5 17. dxe5 Nc5 18. Nd4 Ne4 19. Kg2 Bc5 20. Be3 #dia


click for larger view

I obviously chose these most defensive CK lines for Black, because it gives White the opportunity for a kingside attack.

However, SF says -0.0 (unclear), we used about 2 hour of analysis for 7 moves (13-19).

May-24-18  WorstPlayerEver: PPS Something funny happened,

20. Be3 Nxg3 21. hxg3 Qe8 22. Rc1 exf5 23. Rxf5 Rxf5 24. Nxf5 Bxe3 25. Nxe3 Qxe5 26. Nxd5 #dia

-While 26. Nf5 with advantage for Black.


click for larger view

Black to move..

May-24-18  WorstPlayerEver: PPPS

-22. Qd3 exf5 23. Nf5 Bxe3 24. Qxe3 Ne6 25. Qd3 Rd8 26. Rac1 Nc5 27. Qd4 Qd7 28. Qd1 Qd7 29. Qd3 Nc5 30. Qd4 Ne6 1/2-1/2

May-25-18  WorstPlayerEver: Tremendous fight. Leela sacced a Knight at f5, got 3 pawns. Miraculous save by SF. Leela kept pushing for 30 moves. SF evaluated some positions +1. One position I kept the engines running for 8 hours. Position after 36 moves (White to move):


click for larger view

Position = 0.00 however, if Leela finds another improvement in the first 20 moves, then the iron defense CK variation could be history...

May-25-18  WorstPlayerEver: 27...Qxg5 0-1 #dia

Super Caro-Kann! Leela 338 tried to develop the DSB c1 to e3. Without being too fruitful.


click for larger view

May-25-18
Premium Chessgames Member
  Count Wedgemore: <WorstPlayerEver: 27...Qxg5 0-1 #dia--

--Super Caro-Kann!>

I knew it! After this, who will dare mocking us Caro-Kann players?

Seems like we've been right all along, huh? LOL! Great work, by the way. An interesting study you've been doing.

May-27-18  WorstPlayerEver: <Count Wedgemore>

Yes, it has been a great experiment. Because Leela never plays the same opening twice.

WPC (1st game White)-Leela 346

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. e3 b6 5. Ne2 O-O 6. a3 Be7 7. e4 d5 8. cxd5 (TN) exd5 9. e5 Ne4 10. Nxe4 dxe4 11. Nc3 Bb7 12. Bc4 Nc6 13. Be3 Bg5 14. O-O h6 15. Bd5 Bxe3 16. fxe3 Qg5 17. Qe2 Rab8 18. Nxe4 Qg6 19. Nf6+ 1-0


click for larger view

Compare:

G Nakhbayeva vs M Lubbe, 2010

May-27-18  WorstPlayerEver: PS great sense of humor as well:

19. Nf6+ Kh8 20. Be4 Qg5 21. Rf5 Qh4 22. Rh5 Nxd4 23. exd4 Qf4 24. g3 Ba6 25. Qe1 Qf1+ 26. Qxf1 Bxf1 27. Rxf1


click for larger view

May-27-18  WorstPlayerEver: Second game White. Wherein we give a demo of how well planned dirty tactics well.. uh.. work :)

WPC-Leela 347

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. e3 b6 5. Ne2 O-O 6. a3 Bxc3+ 7. Nxc3 Bb7 8. d5 c6 9. d6 (TN) c5 10. Ra2 Ne8 11. b4 f5 12. b5 Qf6 13. Rc2 Qe5 14. Bd3 #dia1


click for larger view

9. d6 was a risky one. Implies a pawn sac. Leela takes the wrong pawn though. I know it was tempting :)

-14. Bd3 Bxg2 15. f4 Qf6 16. Rxg2 Qxc3+ 17. Rd2 a6 18. Bb2 Qa5 19. Rg1 1-0 #dia2 NB Black has not lost a pawn.. yet.


click for larger view

The White Rooks may be the stars of the show, but pawn d6 is its choreographer.

-19. Rg1 axb5 20. cxb5 Rf7 21. Kf1 Qa4 22. Qh5 c4 23. Bxg7 #dia3 Nxg7 24. Rxg7+ Rxg7 25. Qe8#


click for larger view

Jun-03-18
Premium Chessgames Member
  Penguincw: Another AlphaZero vid: https://www.youtube.com/watch?v=CX4....
Jun-15-18  WorstPlayerEver: Leela Zero does not play 1. e4 any longer. I just checked this 25 times.

Can someone confirm this?

Jun-15-18  WorstPlayerEver: Leela ID 411-WPC

Teasing Leela. A curiosity.

1. c4 e5 2. g3 Nf6 3. Bg2 Bd6 4. Nc3 Bb4 5. e4 Nc6 6. Nge2 Bc5 7. h3 d6 8. d3 a5 9. O-O O-O 10. Kh2 Bd7 11. f4 Nd4 12. Nxd4 Bxd4 13. f5 h6 14. h4 c6 15. g4 Nh7 16. Kg3 Kh8 17. Ne2 Bc5 18. Nc3 Rg8 19. Qe2 a4 20. Rb1 Bd4 21. Bd2 a3 22. b3 g6 23. Kh3 h5 24. g5 f6 25. Kh2 fxg5 26. hxg5 gxf5 27. exf5 Bxc3 28. Bxc3 Qxg5 29. Be1 Bxf5 30. Rf2 Qg3+ 31. Kh1 Bxd3 32. Qxh5 Bxb1 33. Re2 Qg4 34. Qxg4 Rxg4

Jul-09-18
Premium Chessgames Member
  AylerKupp: I haven't visited this page in a while and it seems that it has turned into a Leela Chess Zero (LCZero) page rather than an Alpha Zero page. Oh well, I suppose it was inevitable.

I've been meaning to post on an interesting development. In AlphaZero - Stockfish (2017) (kibitz #144) I said that "I also think that it would be relatively straightforward for any engine to replace their hand crafted evaluation function with a simulation game approach as used in AlphaZero." but I thought that the most likely candidate would be Stockfish since it's code is in the public and there might be quite a few programmers that would be interested in teaming up and making that modification. I also said that "It would be<VERY> interesting to see how a version of Stockfish with such a probabilistic approach to its evaluation function would do against the same version of Stockfish with its "classical" evaluation function and with both versions running on the same hardware configuration."

Well, it has come to pass but with Komodo instead of Stockfish. Since early June Komodo 12 has been shipping with an option to use MCTS instead of alpha-beta pruning to determine a position's evaluation. And it didn't take all that much time, about 6 months since the publication of the initial AlphaZero paper in Dec-2017 and the release of Komodo 12 MCTS in June-2018. So I still think that it would be <VERY> interesting to see how Komodo 12 "classical" performs in a match with Komodo 12 MCTS when running on the same hardware. It's not obvious by looking at their lists that Komodo 12 MCTS is being used in either the CCRL 40/40 or CEGT 40/20 engine tournaments.

Maybe I'll try to do that once I purchase Komodo 12.

Jul-10-18  nok: <Apr-19-18 nok: Open a page for Leela, cg.>

AlphaZero (Computer) (kibitz #178)

Being an open learning project anyone can contribute to, Leela is more important than Alpha. But cg's management is a bit slow on the draw.

Jul-10-18
Premium Chessgames Member
  AylerKupp: <<nok> Leela is more important than Alpha. >

I agree. Now that Google/Deep Mind has gotten the publicity they wanted from AlphaZero I doubt that you will hear from it again, although it's technology might find its way into other products. Similar to what happened with Deep Blue in 1997 when IBM decided that it had achieved what it wanted to achieve publicity-wise and so they dismantled it.

Jump to page #    (enter # from 1 to 9)
search thread:   
< Earlier Kibitzing  · PAGE 9 OF 9 ·  Later Kibitzing>
NOTE: You need to pick a username and password to post a reply. Getting your account takes less than a minute, totally anonymous, and 100% free--plus, it entitles you to features otherwise unavailable. Pick your username now and join the chessgames community!
If you already have an account, you should login now.
Please observe our posting guidelines:
  1. No obscene, racist, sexist, or profane language.
  2. No spamming, advertising, or duplicating posts.
  3. No personal attacks against other members.
  4. Nothing in violation of United States law.
  5. No posting personal information of members.
Blow the Whistle See something that violates our rules? Blow the whistle and inform an administrator.


NOTE: Keep all discussion on the topic of this page. This forum is for this specific player and nothing else. If you want to discuss chess in general, or this site, you might try the Kibitzer's Café.
Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors.
Spot an error? Please suggest your correction and help us eliminate database mistakes!


home | about | login | logout | F.A.Q. | your profile | preferences | Premium Membership | Kibitzer's Café | Biographer's Bistro | new kibitzing | chessforums | Tournament Index | Player Directory | Notable Games | World Chess Championships | Opening Explorer | Guess the Move | Game Collections | ChessBookie Game | Chessgames Challenge | Store | privacy notice | contact us
Copyright 2001-2018, Chessgames Services LLC