Originally posted by Paul Bonham
Originally posted by Paul Bonham
-
Sure, and "for all we know" the Apollo mission to the moon was shot in a movie studio.
That's fine, but if you have that much of an aversion to chess opening theory and even chess theory in general then it underscores my point that your commentary on Stockfish's choice of lines in an opening is of little relevance. I do not want to play option chess for exactly the opposite reason.Last edited by Sid Belzberg; Friday, 28th December, 2018, 10:56 AM. -
For all we know, the programmers of AZ could have inserted code in the engine to say that "if you find an early opening move that is very rare but at the same time is about equally strong as the usual move, play the rare move to get the opponent out of book or out of familiar territory" and AZ could have connections to chess databases to find out which moves fit that category. The 4....a6 move definitely fits the bill, as it has been very rarely played and yet SF8 itself considers it almost equal with 4...Nf6.Originally posted by Sid Belzberg View Post
If the above were to be the case, there is absolutely nothing otherworldy about it. Just clever programming.
The other point I would make is that we don't yet know how often AZ is willing to play these "novelties". Have it play 10,000 games of the Giucco Piano as Black up to move 4.d3 and see how many times it plays 4...a6 versus 4...Nf6. If the 4...a6 move predominates, then either the programming effect I noted above is in effect OR AZ really has found something about 4...a6 by studying complete game results in which the games are generated internally by semi-random move selection.
I have no problem admitting that AZ plays very different stylistically from all other chess engines (excepting LCZero which is modeled after AZ). For that alone, we should all be glad AZ has come along and we should hope many more similar engines can be available to use on typical computer hardware very soon. This would get people studying computer played chess much more than in the past. Computer chess has developed a reputation for being very dry and uninteresting to most.
Sid I do not want to play chess960 as it devolves into regular chess much too quickly. I haven't played competitive chess at all in over 20 years. Option Chess only has short segments where it may become like regular chess, so the entire game throughout is different. I'll leave my challenge open. You really have very little to lose and a lot to gain -- if you took time away from playing regular chess on that app you mentioned and spent it instead on playing a much longer more involved game of Option Chess, you'd be a better person for the experience imo.
I actually thought you were much higher rated than 2032, I thought you were up around 2400 or at least 2300. Maybe at some time you were.Leave a comment:
-
Yes, I was also impressed by that. I was actually surprised that this move was tried at all in any serious game. I was equally surprised that no human in any of the cited games had ever followed up with the move Stockfish did 5 N-g5. The discussion sort of takes away from the main point I was making that we are seeing many new novel ideas like 4a6 that as we can see up to now are rarely considered and indeed the play of Alphazero has been described by many as "other worldly" and I absolutely agree with this characterization.Originally posted by Tom O'Donnell View Post
Already similar methodologies are being deployed for things like near instant diagnosis for possible stroke by comparing a patients CAT scan to large databases of known CAT scans. I consider this technological breakthrough to be almost on the same order of importance as the elucidation of the DNA molecule in 1953.Leave a comment:
-
I will admit my bias upfront that reading this thread I didn't think 4...a6 demonstrates much of anything, though of course asking one move to prove anything is a pretty small sample size. However, I am really impressed with the amount of research that went into this post.
Originally posted by Paul Bonham View PostLeave a comment:
-
That is a simply a rationalization of your own shortcoming in learning chess. If you want fast and easy proof of that download "social chess" on your phone and challenge me to "chess960" which is what they call FischerRandom chess. My handle is Sidney 81934. i play lots of chess960 mixed in with normal chess and my rating there is not much different then my USCF and Chess.ca ratings, (2032 at this particular moment.)I play at a time control of 5 0.Originally posted by Paul BonhamLast edited by Sid Belzberg; Thursday, 27th December, 2018, 02:08 PM.Leave a comment:
-
Originally posted by Sid Belzberg View Post
Actually, what I am wondering is whether 5.Ng5 as played by Stockfish8 after 4....a6 was not a calculated move at all, but was instead picked out of SF8's opening book.
With my version of SF8, I do not have it using any opening book. So when I play the opening 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.d3 a6 and then ask SF8 to give me it's top 6 moves in reply, here is what it gives me after a 31-ply search depth:
1 31 +0.23
5.a4 Nf6 6.O-O O-O 7.c3 Ba7 8.Re1 d6 9.Nbd2 Ng4 10.Rf1 Ne7 11.h3 Nf6 12.Re1 Ng6 13.Nf1 c6 14.Bb3 Re8 15.Ng3 Be6 16.d4 exd4 17.cxd4 Qb6 18.Bxe6 Rxe6 19.b4 Rae8 20.Be3 d5 21.e5 Ne4 22.Nxe4 dxe4 (1339.05)
2 31 +0.20
5.O-O Nf6 6.c3 d6 7.b4 Ba7 8.Bg5 O-O 9.Nbd2 h6 10.Bh4 g5 11.Bg3 b5 12.Bb3 a5 13.a4 axb4 14.axb5 bxc3 15.bxc6 cxd2 16.Qxd2 Nh5 17.d4 Bg4 18.dxe5 Nxg3 19.hxg3 Bxf3 20.gxf3 dxe5 21.Qxd8 Rfxd8 22.Ra5 Bb6 23.Rxe5 (1339.05)
3 31 +0.19
5.Ng5 Nh6 6.c3 O-O 7.a4 d6 8.h3 Kh8 9.b4 Ba7 10.O-O Qe8 11.Qh5 Bd7 12.Na3 Ne7 13.a5 Ng6 14.Nc2 f6 15.Nf3 Ng8 16.Nh4 Nxh4 17.Qxh4 Ne7 18.Be3 Bxe3 19.Nxe3 f5 20.exf5 Nxf5 21.Nxf5 Rxf5 22.Bd5 Rf4 23.Qg3 Bc6 24.Bb3 h6 (1339.05)
4 31 +0.19
5.c3 Nf6 6.a4 d6 7.b4 Ba7 8.O-O O-O 9.Re1 Ne7 10.Bb3 Ng6 11.a5 h6 12.Be3 Bxe3 13.fxe3 Re8 14.Qc2 Bd7 15.Nbd2 Qe7 16.Rf1 Rac8 17.h3 Ra8 18.Rab1 Rad8 19.Rbe1 c5 20.Nc4 cxb4 (1339.05)
5 31 +0.14
5.h3 Nf6 6.O-O O-O 7.c3 Ba7 8.Re1 d6 9.Bb3 Re8 10.Nbd2 h6 11.Nf1 Bd7 12.a4 Na5 13.Ba2 c5 14.Ne3 b5 15.axb5 axb5 16.Bd5 Nxd5 17.Nxd5 Nc6 18.Be3 Ne7 19.Qb3 Nxd5 20.Qxd5 Qf6 (1339.05)
6 31 +0.10
5.Nbd2 Nf6 6.O-O O-O 7.a4 Ba7 8.b4 Ne7 9.Bb3 d6 10.c3 Ng6 11.Nc4 Be6 12.Re1 h6 13.a5 Nh5 14.Be3 Nhf4 15.Bxa7 Rxa7 16.d4 exd4 17.cxd4 Nh4 18.Nxh4 Qxh4 19.Qf3 Re8 20.h3 d5 (1339.05)
Notice how close the top 4 moves are: varying by only 0.04 points. The 5.c3 move that has been favored by humans is tied with 5.Ng5 in SF8's estimation. I really believe SF8 chose 5.Ng5 out of an opening book, without even doing any calculation.
But the whole point is this: 4....a6 and even 5.Ng5 did NOT decide this game! Sid, you are talking as if AZ won the game simply by playing 4...a6 and that is rubbish. I went to the Talkchess.com site (a site for chess programmers) to see what was being said about this game there, and the big talk was about 13....Ne7. There was not a word about 4...a6.
Perhaps some tournament organizer should hold a theme tournament with exactly this opening up to 4....a6 and gather together the strongest players possible to play all games with this opening, and we'd have a much better handle on the situation. Why don't you finance that Sid? Put your money where your mouth is! If 4...a6 is really some sort of super move in that variation, then such a tournament should have unusually good results for Black. Then you would really have some evidence for your claims.
I wonder if it has occurred to you that the rarity of 4...a6 in the GP opening says more about human players than it does about AZ. I'll grant you that AZ is the strongest player to ever play that move, that's obvious, but I'll not grant you that AZ has somehow "proven" that 4...a6 is the very best move in that opening. The human games show that 5.c3 is a very daunting response.
Your comment about my "B player strength chess wisdom" is very insightful. You seem to imply that I can't reason very well about anything to do with chess because I was only a B strength player.
I wonder Sid if you would be up to a challenge. I'd like to play you a 2-game match of correspondence Option Chess, a game of my invention. Now before you rush to type a knee-jerk response, let me assure you that there is no computer engine in the world that can play Option Chess, so neither you nor I would have any computer help. Even I as the game's inventor have no special insight into how to play Option Chess in any way, and in fact the conditions of the match would be that we each get to choose the first 16 plies of the game ( which are played as normal chess) in 1 of the 2 games. I have done no studies of any kind in "opening theory" of Option Chess, for that you have my word. So the match between us would be even, and your superiority to me at regular chess should actually give you a substantial edge.
I have played only 4 games of Option Chess myself, all against opponents who are much better at regular chess than I am. My result is 3 losses and 1 draw. So there you go, you should win such a match easy!
Well, except for maybe one thing. In those other 4 games I played, I was playing in a deliberate "experimental" mode, not really caring much about winning or losing but rather about discovering something. Against you, I will care only about winning.
My point will be to show this: that even with my significant inferiority to you in regular chess, due to things like what age we each started playing and how much time we put into it, I can hold my own against you in a game that is very much like regular chess BUT has no opening theory nor established theory of any kind. The moves are the same, the board and pieces are the same, but the game is different. And you will have as much time as you like to study the rules of the game. I'll even supply you with the 4 games I've played so far, with full annotations.
I'm open to suggestions as to what the stakes of the match would be. And I'll go one step further: in order for me to win the match, I must score 1.5 points out of 2. That is to reassure you in case you think I might (as the game's inventor) have some unknown edge which I assure you I don't. If I were to accomplish such a feat, I'd not ask much of you. Maybe just to finance an Option Chess weekend over-the-board tournament at a club close to you, something like that.
So I'm putting out the challenge and the odds are greatly in your favor. We would each have one full week to make a move, and could in case of unforseen circumstances have 2 "time outs" of up to a month. So the match could take a year or longer to finish, but what a year that would be! lol
I know you like challenges, especially intellectual ones. Have a go at it!
And by the way, this does have pertinence to the discussion about AZ. Because it is my theory that AZ would not dominate humans at Option Chess the way it does at regular chess. For that to be proven, we first need to produce some good human Option Chess players.
Leave a comment:
-
Really???, You are questioning Stockfish's choice of moves, a machine that is far stronger then any human but far weaker then Alpha Zero. On what basis? That no human ever replied with N-g5 in the very few games that this move a6 ever showed up? Or is this a pearl of your B player strength chess wisdom?Originally posted by Paul Bonham
By the way, not one GM ever took up this move up so indeed a6 is a novelty as it would be considered just that if a strong human super GM took it up today. The fact is that AlphaZero is stronger then any human and is the first "super GM" strength program to introduce this move in recorded play and beat the worlds strongest program Stockfish with it. Yes, indeed that is novel and is but one example of many amazing novel chess ideas introduced by Alpha zero.
Even after 30 ply Stockfish did not rank a6 as the best move and then got trounced by AlphaZero with just that move.Last edited by Sid Belzberg; Wednesday, 26th December, 2018, 12:25 PM.Leave a comment:
-
So now you are even going so far as to say 4...a6 is not just a novelty but is the best move. Well I have some news for you.Originally posted by Sid Belzberg View Post
I did a rudimentary Google search of the Giucco Piano opening move sequence up to 4...a6 and came up with several hits on the chesstempo.com web site, which is an online database of about 2 million games. I found the following 8 games listed below. In all 8 games, Black is rated over 2000 ELO.
The very first one is a game from 1988 where GM Vlastimil Hort is White, and his opponent is rated 2335. So that blows your whole theory about never seeing 4...a6 at that level in our lifetimes.
I also draw your attention to the 5th game, Hofmann, Lorenz (1663) vs Logar, Metod (2024). Black has over 300 ELO rating advantage and plays 4...a6 and is held to a draw. The 5.c3 response seems to be very powerful.
Also in all 8 games, White did NOT respond with the move you assume would be always forthcoming in response. Instead of your 5.Ng5 that you seem to think is automatic, the move 5.c3 was played in every game below.
Your talk about 4...a6 as some kind of miracle move that no human would ever think to play and that shows AZ is some kind of other worldy chess intelligence is reduced to rubbish. You were hoodwinked by the person who analyzed that game on YouTube, whoever he is. The real question of that game is why does Stockfish 8 reply 5.Ng5 which which may not be nearly as good as it looks.
The truth about 4....a6 is that Stockfish 8 itself ranks it the second strongest move in the position, as I found out by running it to 30 ply search depth:
1 30 +0.16
4.... Nf6 5.c3 a5 6.O-O d6 7.a4 h6 8.h3 O-O 9.Re1 Be6 10.Bxe6 fxe6 11.Be3 Bxe3 12.Rxe3 Nd7 13.Qb3 Nc5 14.Qb5 Ra6 15.Nbd2 Qf6 16.Qc4 Nd7 17.Ree1 Nb6 18.Qb5 Qg6 19.d4 exd4 20.cxd4 (1107.04)
2 30 +0.23
4.... a6 5.Ng5 Nh6 6.a4 d6 7.h3 O-O 8.O-O Kh8 9.c3 Qe8 10.b4 Ba7 11.Qh5 Ne7 12.Bd2 Bd7 13.a5 Ng6 14.Na3 f6 15.Nf3 Nf7 16.Nh4 Nxh4 17.Qxh4 Be6 18.Rfb1 Kg8 19.Bxe6 Qxe6 20.b5 d5 21.b6 cxb6 22.exd5 Qxd5 (1107.04)
As you can see, SF8 has 4....a6 only 0.07 points worse than 4....Nf6 which is negligible. So much for no GM recommending this move, it is as playable as 4...Nf6 and this according to SF8 not AZ.
There is nothing about 4....a6 that is revolutionary. I have to wonder whether AZ would even play this move repeatedly in that opening, or would it only play it sporadically or maybe 50% of the time with the other 50% being 4...Nf6. The truth may be that in playing more like a human, AZ is mixing up its repertoire and not really telling us with each "novelty" it plays that "this move is actually much better than you humans have ever realized". The real message may be "I'm playing like a human and changing up my opening moves quite often compared to an engine that uses a rigid opening book."
Here are the 8 games I found:
Hort, Vlastimil (2590) vs Bastian, Herbert (2335)
Date: 1988
Event: Bundesliga 8788, Bundesliga 1987/88, Germany
Round: 1
Result: ½-½
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. O-O Ba7 7. Re1 d6 8. Bb3 h6 9. Nbd2 g5 10. Nf1 g4 11. N3d2 Nh5 12. Nc4 Qf6 13. Be3 Nf4 14. Bxa7 Rxa7 15. Nce3 Ne7 16. d4 h5 17. Nd5 Nexd5 18. exd5 O-O 19. Ng3 Ra8 20. Qd2 Bd7 21. dxe5 dxe5 22. Ne4 Qg7 23. Nc5 Bf5 24. Bc2 Bxc2 25. Qxc2 b6 26. Nd3 Qg6 27. Rad1 Rfe8 28. c4 Rad8 29. Qc3 Nxd3 30. Rxd3 a5 31. g3 e4 32. Rd4 f5 33. c5 h4 34. gxh4 bxc5 35. Qxc5 Qf6 36. h5 Qe5 37. Rc4 Re7 38. Qxa5 Rxd5 39. Qa6 e3 40. Rxe3 Rd1+ 41. Kg2 Qd5+ 42. Kg3 Rxe3+ 43. fxe3 Rg1+ 44. Kf4 Qf3+ 45. Kg5 Qxe3+ 46. Rf4 g3 47. Qc8+ Kf7 48. Qxf5+ Ke8 49. Qe4+ Qxe4 50. Rxe4+ Kf7 51. h4 Rb1 52. Rg4 Rxb2 53. Rxg3 Rxa2 54. Kh6 Ra4 55. Kh7 Rxh4
Varacalli, Francisco (2285) vs Benedetti, Julio (2335)
Date: 2018-04-19
Event: 3rd ITT Magistral IM 2018, Buenos Aires ARG
Round: 1.5
Result: ½-½
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. Nbd2 d6 7. Bb3 Ba7 8. Nf1 Be6 9. Ng3 h6 10. O-O Qd7 11. Bc2 Bg4 12. Be3 Nh5 13. Bxa7 Rxa7 14. h3 Bxf3 15. Qxf3 Nf4 16. Nh5 Nxh5 17. Qxh5 g6 18. Qe2 g5 19. Rad1 Ne7 20. d4 Qb5 21. Qe3 Ra8 22. Bb3 Ng6 23. dxe5 Qxe5 24. g3 Qe7 25. f4 gxf4 26. gxf4 Rg8 27. Kh2 Kd7 28. e5 Rae8 29. Qa7 Rb8 30. Ba4+ c6 31. Rxd6+ Kc8 32. Bxc6 Nxe5 33. Qc5
Sahl, Sheila Barth (2186) vs Maroroa, Sue (2102)
Date: 2018-03-18
Event: 4NCL 2017-18, England ENG
Round: 8.28
Result: ½-½
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. b4 Ba7 7. O-O d6 8. Nbd2 Ne7 9. Re1 Ng4 10. Re2 Ng6 11. Nf1 O-O 12. h3 Nf6 13. Ng3 Be6 14. Bb3 Qd7 15. Bxe6 fxe6 16. d4 exd4 17. cxd4 e5 18. dxe5 dxe5 19. Qxd7 Nxd7 20. Rd2 Nb8 21. Bb2 Nc6 22. a3 Rae8 23. Rd7 Bb6 24. Rad1 Re7 25. Nf5 Rxd7 26. Rxd7 Rf7 27. Rxf7 Kxf7 28. Ng5+ Kg8 29. Ne6 Kf7 30. Ng5+ Kg8 31. Ne6 Kf7 32. Ng5+ Kg8
Sypien, Mateusz (2228) vs Papista, Akos (2096)
Date: 2017-03-29
Event: 28th NATO Chess 2017, Budapest HUN
Round: 4.13
Result: 1-0
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. O-O d6 7. Bb3 Ba7 8. h3 O-O 9. Bg5 Be6 10. Nbd2 Qe7 11. Nc4 h6 12. Bd2 Rad8 13. a4 Nd7 14. Ne3 Bxe3 15. fxe3 Bxb3 16. Qxb3 Nc5 17. Qc2 d5 18. exd5 Rxd5 19. d4 e4 20. Ne1 Nd7 21. Qb3 Nb6 22. c4 Rg5 23. a5 Nd7 24. Qxb7 Rg6 25. Qxc7 Nxd4 26. exd4 e3 27. Bc3
Hofmann, Lorenz (1663) vs Logar, Metod (2024)
Date: 2015-08-16
Event: 24th Feffernitz Open 2015, Feffernitz AUT
Round: 1.20
Result: ½-½
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. h3 Nf6 6. Bg5 h6 7. Bh4 d6 8. Nc3 Be6 9. Bb3 g5 10. Bxe6 fxe6 11. Bg3 Qe7 12. Qd2 Nh5 13. Bh2 O-O-O 14. O-O-O Rhf8 15. Qe2 Qf7 16. Bg3 Nf4 17. Bxf4 Qxf4+ 18. Kb1 h5 19. Qd2 Rg8 20. Rdf1 Rdf8 21. Qxf4 Rxf4 22. Nd1 Nd4 23. Nxd4 Bxd4 24. f3 Rgf8 25. Re1 Kd7 26. c3 Bb6 27. Kc2 Ke7 28. Kd2 Kf7 29. Ke2 Kg6 30. Ne3 c6 31. Rhf1 Bxe3 32. Kxe3 g4 33. hxg4 hxg4 34. Rf2 Kg5 35. Rh1 R8f7 36. Rh8 Rf8
Gulbas, Cemil (2371) vs Lenaerts, Lennert (2175)
Date: 2013-12-15
Event: Wirtzfeld - CRELEL
Round: 6.3
Result: 1-0
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. Nbd2 Ba7 7. Bb3 d6 8. Nf1 h6 9. Ng3 d5 10. Qe2 O-O 11. O-O Re8 12. h3 Be6 13. Bc2 b5 14. Rd1 Qe7 15. Nh4 Qf8 16. Qf3 dxe4 17. dxe4 Nh7 18. Nhf5 Ne7 19. Nh5 Bxf5 20. exf5 f6 21. Qg4 Kh8 22. a4 Ng8 23. Be4 Rad8 24. Rxd8 Rxd8 25. axb5 a5 26. Rxa5 Bb6 27. Ra1 Ng5 28. Bc2 Rd7 29. h4 Nh7 30. Bb3 Rd8 31. Be3 Bxe3 32. fxe3 Rd7 33. Be6 Re7 34. Qe4 Qd8 35. Qa8 Re8 36. Qxd8 Rxd8 37. Ra7 Nf8 38. Bxg8 Rd1+ 39. Kh2 Kxg8 40. Rxc7 Rd7 41. b6
Codenotti, Marco (2063) vs Van Hoolandt, Patrick (2183)
Date: 2008-04-26
Event: XIV Leonardo di Bona Magistrale, Cutro ITA
Round: 4
Result: 0-1
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 a6 5. c3 Nf6 6. Bb3 O-O 7. Nbd2 Ba7 8. h3 h6 9. Nf1 d5 10. Qe2 Be6 11. g4 dxe4 12. dxe4 Bxb3 13. axb3 Qd6 14. g5 hxg5 15. Bxg5 Nh7 16. Rd1 Qe6 17. Rg1 f5 18. Ng3 fxe4 19. Nxe4 Rf5 20. Nh4 Bxf2+ 21. Nxf2 Rxg5 22. Rxg5 Nxg5 23. Qh5 Nh7 24. Ne4 Re8 25. Nc5 Qf7 26. Qf5 Qe7 27. Qf2 Nd4 28. Ne4 Qxh4 29. cxd4 Qxe4+
Mouradian, Knarik (2119) vs Ortiz, Nadya Karolina (2148)
Date: 2004-10-24
Event: 36th Olympiad w, Calvia ESP
Round: 9
Result: 0-1
Opening: Italian Game, Giuoco Pianissimo (C50)
1. e4 e5 2. Bc4 Bc5 3. Nf3 Nc6 4. d3 a6 5. c3 Nf6 6. O-O Ba7 7. Bb3 d6 8. Nbd2 O-O 9. Re1 h6 10. Nf1 Nh5 11. h3 Qf6 12. Ne3 Nf4 13. Ng4 Bxg4 14. hxg4 g5 15. Be3 Ne6 16. Bxa7 Rxa7 17. g3 Kg7 18. a3 Rh8 19. Kg2 Raa8 20. Rh1 Rh7 21. Qe2 Rah8 22. Rh5 Kf8 23. Rah1 Qg6 24. Bxe6 fxe6 25. Qe3 Ke8 26. b4 b6 27. a4 a5 28. b5 Nb8 29. Nd2 Nd7 30. f3 Nf6 31. R5h2 h5 32. gxh5 Nxh5 33. Qe2 Qf7 34. Qe3 Rg8 35. Qe2 Nxg3 36. Kxg3 g4 37. Rxh7 gxf3+ 38. Kf2 Rg2+ 39. Kf1 fxe2+ 40. Kxg2 Qxh7 41. Re1 Qh5 42. Nf3 Qh6 43. Kf2 Kd7 44. Kxe2 c6 45. c4 Qh8 46. Rc1 Qc8 47. Kd2 Qf8 48. Rf1 Qf4+ 49. Ke2 cxb5 50. cxb5 d5 51. Nd2 Qg4+ 52. Rf3 Kd6 53. Nb3 dxe4Leave a comment:
-
A "bit later in the game" is very different from move 4. 4 a6 immediately provokes n-g5 forcing the "ugly" n-h6 and all the complications that follow. Moving a6 a bit later is routine, especially after after n-f6 and 0-0 is normal as it allows an escape square for the bishop on c5 and has other attributes that are more standard and expected such as supporting b7-b5 followed by b-b7 etc. I am surprised that it appears in megabase at all albeit 7.1 million + games is alot of games. How many of the 204 games were simply beginners? Which 2400 rated player played it or is it a computer? When did these 204 games appear and how many of the 204 were at master level chess or even expert level for that matter. Zero games at grandmaster level is no surprise, and that is my point, Alphazero shows that would have never been believed at the highest levels that 4....a6 is in fact the strongest move. It reminds of an April fools joke Martin Gardner published in 1975 in Scientific American claiming a computer discovered 1) h4 to be a winning move. Well it is not April fools and Alpha zero is showing something almost as preposterous with 4)a6 except it is for real!Originally posted by Vlad Drkulec View Post
In any event dozens of weird early strange moves have been already produced by alpha zero that few would have imagined are the best move, in particular at the GM level and 4 a6 is one of many interesting novelties at this elevated level of chess.
As for discouraging your students from playing this move, I am sure you are in good company as until now likely any GM in the world would have given the same guidance.Last edited by Sid Belzberg; Tuesday, 25th December, 2018, 03:26 PM.Leave a comment:
-
Actually, the move appears 204 times in Megabase 2019 by players rated as high as 2400. The move also independently appeared in the games of at least three or four of my beginner students in the last few months and a few more over the last few years. I have discouraged them from playing the move this early. This may be a clear sign of my lack of imagination. I have seen it played a bit later in the games of even 2800 or 2700 players.Originally posted by Sid Belzberg View PostLeave a comment:
-
Our dear former FIDE President, Kirsan Ilyumzhinov, always has claimed that chess itself is extra-terrestrial!
He was advised by the aliens that a long time ago, when they visited earth, they gave their game to various groups of humans in different parts of the globe (At that time geography was a great barrier to movement). The various nations then adapted the game according to their own culture, time and beliefs about the game......thus we end up with different varieties of chess being developed in different parts of the world.
I wonder if it is the aliens who have recently introduced humans to AI? Might explain why we are somewhat in awe and perceive that AI is extra-terrestrial as well.
Bob ALeave a comment:
-
For anyone wondering about the difference between so-called "minimax" engines and mcts, minimax does a search only to a very shallow (in chess terms) depth, and then uses an "evaluation function" to try and determine the value of the position, then feeds that value back. This evaluation function is where the difference lies in various engines. For example, Stockfish's evaluation function might value Pawn structure ahead of passed Pawns, whereas another engine's might not. The evaluation function implies the possibility of error. Whereas with mcts, games are played to full completion in memory using random or heuristical move generation, and the actual game results are fed back. In other words, no evaluation function. It is still possible for errors -- the move generation might miss a corner case winning line. But generally speaking, the mcts method is more human like because humans learn over time by playing complete games. When you're a junior and losing games, you learn things. Mcts engines do this same process in memory (which might be in the Terabytes on a distributed compute cluster) except they do it millions of times faster and with perfect feedback.
You say that the human approach to chess is more akin to the MCTS method because that is what juniors do when learning the game. But it seems to me that juniors after playing some large number of games develop a 'feel' for a type of position - or rather an evaluation for such things as mobility, piece strength, doubled pawns etc. This therefore is more akin to an 'evaluation' function. However many games a junior may play during the learning stage it is not comparable to the MCTS where gazillions of games would be played and remembered.
Therefore, I have to side with the argument that Alphazero plays otherwordly style chess. The term I used in my initial post when I started this thread was extraterrestrial and I still stand by that.Last edited by Vlad Dobrich; Monday, 24th December, 2018, 12:19 PM.Leave a comment:
-
Yes my point exactly, "Eventually", in this case I very much doubt if any master strength player would come up with this in our life time. I will say that in another version of the Italian game, the two nights defense, Fischer in 1963 playing Bisguer revived a novelty first introduced in the 19th century by Steintz 9.n-h3 setting up a very similar strategic theme with the white pieces as Alpha Zero was aiming for with black in the above example. But, I digress. The point is that a machine that goes through trillions of patterns and finds the best ones does produce an array of novelties at a rate the likes of which have never been seen before on this planet.Originally posted by PaulBonham
As mentioned in an earlier thread we are already seeing some limited success in using this methodology in games of incomplete information like Poker. In heads up competition the machine routinely sends the top players to the rail. In multi player tournament play it is more problematic probably because the machine does not have enough of its opponents hand history to learn it's opponents patterns so with best play it is more or less a coin flip although I suspect the machine has a slight edge as it does not "tilt" and can more accurately remember its opponents strategies from what little information it can glean.
https://www.youtube.com/watch?v=G5519_PzT8gLast edited by Sid Belzberg; Monday, 24th December, 2018, 12:21 PM.Leave a comment:
-
Originally posted by Sid Belzberg View Post
I suppose we can each of us have our own definition of "other worldly" so I won't belabor the point. To a Neanderthal, the steam engine would be other worldly. To us in today's world, it is somewhat antiquated.
I contest your point that no engine or serious chessplaying human being would ever have considered 4...a6 in the GP opening. The fact that it hasn't been considered so far doesn't mean it would never have happened in any future timeline. It could just be that the serious chessplayers of the last few decades have been working on other openings, not even paying much attention to the GP, or that in studying the GP they've just been focusing on later moves. The point is, EVENTUALLY someone would have gotten around to 4....a6. Humans do after all discover opening novelties.
So is it other worldly that AlphaZero is the first to consider this early move? Well, to some it could be. But remember that AZ is by its very nature a human-like engine. The only difference is that it is supported by a massive technological infrastructure, the bleeding edge of AI technology. That immediately means that it plays like a human but both remembers and calculates much much better.
The fact that it plays like a human has only been enabled by the advances in technology. If you consider the progression of chess engines to the present day, it's all very logical and orderly. Minimax engines dominated for decades precisely because there wasn't the technology to support the idea of monte carlo tree search (mcts)... the monto carlo method existed but the required technology to make mcts feasible didn't. The term "monte carlo tree search" wasn't even coined until 2006.
The very idea of playing full games from a given position, storing the result of each, feeding back those results and doing this enough times to actually learn something just wasn't feasible in 1980 or 1990 or 2000. MCTS is an idea whose time has come, thanks to the very latest advances in GPGPU (General Purpose Graphics Processing Unit) technology. Sid if you disagree with this, ask yourself why up until very recently -- literally the last few months -- it took millions of dollars to support an AlphaZero engine? You gave the example of LCZero which can now apparently run on GPUs costing a few thousands of dollars. In 10 years, we might all have an AZ that never loses at chess built into our HUD (Head Up Display) which all of us might be wearing by then.
For anyone wondering about the difference between so-called "minimax" engines and mcts, minimax does a search only to a very shallow (in chess terms) depth, and then uses an "evaluation function" to try and determine the value of the position, then feeds that value back. This evaluation function is where the difference lies in various engines. For example, Stockfish's evaluation function might value Pawn structure ahead of passed Pawns, whereas another engine's might not. The evaluation function implies the possibility of error. Whereas with mcts, games are played to full completion in memory using random or heuristical move generation, and the actual game results are fed back. In other words, no evaluation function. It is still possible for errors -- the move generation might miss a corner case winning line. But generally speaking, the mcts method is more human like because humans learn over time by playing complete games. When you're a junior and losing games, you learn things. Mcts engines do this same process in memory (which might be in the Terabytes on a distributed compute cluster) except they do it millions of times faster and with perfect feedback.
To me this isn't other worldly, this is just the natural progression of things. I remember in the late 1990's I investigated a chess engine called Octavius which was a neural net engine. It all seemed very promising, but the technology wasn't there to support it. It played at a sub-2000 ELO level. Nevertheless I was impressed and ever since then I've been expecting a NN style engine to eventually come along and rise to the top given the right advances in technology.
In a non-perfect information game such as poker, an AZ type of engine could only be expected to be a top performer, not doing significantly better or worse than the best human players. It wouldn't beat humans every time as it would in chess. The reason is that in poker, playing the absolutely best play from a probability and statistical point of view is what the best human players have already been doing for decades using well-developed theory, and unlike in chess can still be defeated by the draw of the next card.
And since "the draw of the next card" is what governs most of life around us, an AZ engine within some kind of robot isn't going to eventually run the world to perfection, which I believe goes against the scenario some people are proclaiming. Nevertheless such an AZ robot would certainly help... perhaps such a robot will discover a new way to counter climate change just as AZ "discovered" 4....a6 in the Giucco Piano opening.
If anything is "other worldly" it is the very existence of human beings. Just think about our eyesight and everything that goes into the processing of light rays by our eyes and brains. There's no robot that can even approach what we do naturally. And then there's hearing, and touch, and taste... could all of those actually have happened all at the same time through evolution? They could perhaps have been PERFECTED through evolution, but for each of them to have independently happened from nothing, from a dead universe, seems.... other worldly. If we use that as our standard, there is nothing happening in chess or in technology so far that seems other worldly.
Leave a comment:
-
Yes it is only technology but the results are indeed other worldly. Here is an example of a game with the ancient opening the Giuoco Piano that every beginner loves playing where alphazero plays a bizarre and unique novelty on move 4 (!) 4...a6. No chess engine or serious chessplaying human being would have ever even considered this move. Despite stockfish playing the "best" moves it gets crushed.Originally posted by paul Bonham
Someday Quantum computers will likely be able to resolve a public key back to its derived private key and even though it is only technology this too will seem other worldly and just like this example it is in fact "other worldy".
https://www.youtube.com/watch?v=G-OmM_UWaEgLast edited by Sid Belzberg; Sunday, 23rd December, 2018, 03:25 PM.Leave a comment:
Leave a comment: