Historic Milestone: AlphaZero

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
MikeGL
Posts: 700
Joined: Thu Sep 01, 2011 12:49 pm

Re: Historic Milestone: AlphaZero

Post by MikeGL » Thu Dec 07, 2017 7:38 am

Dirt wrote:I doubt that Google wants to be in the chess playing, or go playing or shogi playing business. I wonder what their goal is?

One possibility is self-driving cars, although they have competition there. Natural language recognition, and possibly translation, is also interesting. Maybe civil engineering?
I think this type of AI is generic, can be applied in StarCraft/WarCraft online to win against humans,
or other similarly sophisticated games. Could predict stockmarket fluctuations more accurately.
Could help in very complex real wars involving jet fighters, submarines and aircraft carriers too.

User avatar
MikeGL
Posts: 700
Joined: Thu Sep 01, 2011 12:49 pm

Re: MCTS-NN vs alpha-beta

Post by MikeGL » Thu Dec 07, 2017 7:59 am

kranium wrote:
Lyudmil Tsvetkov wrote:It is not at all clear to me where were books used and where not.
I'm sure opening books were not used...
In the early self-play games things like 1.a3, 1.a4, etc. were probably tried by AlphaZero...
eventually it learned that 1. e4 or 1. d4 had the highest success rates.
Books or no books, I think AlphaZero would still demolish SF8.
Just look at this game 9, it was a decent French Defence by SF8, but it was dismantled with
amazing tactical and strategic shots by AlphaZero which seems to be beyond the reach of alpha-beta engines.

[pgn]
[Event "?"]
[Site "?"]
[Date "2017.12.06"]
[Round "9"]
[White "AlphaZero"]
[Black "Stockfish"]
[Result "1-0"]
[TimeControl "40/1260:300"]
[Termination "normal"]
[PlyCount "103"]
[WhiteType "human"]
[BlackType "human"]

1. d4 e6 2. e4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 cxd4 7. Nb5 Bb4+ 8.
Bd2 Bc5 9. b4 Be7 10. Nbxd4 Nc6 11. c3 a5 12. b5 Nxd4 13. cxd4 Nb6 14. a4
Nc4 15. Bd3 Nxd2 16. Kxd2 Bd7 17. Ke3 b6 18. g4 h5 19. Qg1 hxg4 20. Qxg4
Bf8 21. h4 Qe7 22. Rhc1 g6 23. Rc2 Kd8 24. Rac1 Qe8 25. Rc7 Rc8 26. Rxc8+
Bxc8 27. Rc6 Bb7 28. Rc2 Kd7 29. Ng5 Be7 30. Bxg6 Bxg5 31. Qxg5 fxg6 32. f5
Rg8 33. Qh6 Qf7 34. f6 Kd8 35. Kd2 Kd7 36. Rc1 Kd8 37. Qe3 Qf8 38. Qc3 Qb4
39. Qxb4 axb4 40. Rg1 b3 41. Kc3 Bc8 42. Kxb3 Bd7 43. Kb4 Be8 44. Ra1 Kc7
45. a5 Bd7 46. axb6+ Kxb6 47. Ra6+ Kb7 48. Kc5 Rd8 49. Ra2 Rc8+ 50. Kd6 Be8
51. Ke7 g5 52. hxg5 1-0
[/pgn]

not sure if 18.g4!, 30.Bxg6! and other would be found by current engines.

[d]r2qk2r/3bbppp/1p2p3/pP1pP3/P2P1P2/3BKN2/6PP/R2Q3R w kq - 0 18
After 17...b6 of black, can some engine consider 18.g4! in this position?



xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
[d]4q2r/1b1kbp2/1p2p1p1/pP1pP1N1/P2P1PQP/3BK3/2R5/8 w - - 6 30
After 29...Be7, can current engines consider 30.Bxg6! here?


Would be nice if we can try to feed some difficult epd positions into AlphaZero,
to estimate its ELO strength.

User avatar
lantonov
Posts: 216
Joined: Sun Apr 13, 2014 3:19 pm

Re: MCTS-NN vs alpha-beta

Post by lantonov » Thu Dec 07, 2017 10:04 am

MikeGL wrote: [d]4q2r/1b1kbp2/1p2p1p1/pP1pP1N1/P2P1PQP/3BK3/2R5/8 w - - 6 30
After 29...Be7, can current engines consider 30.Bxg6! here?
For the second position SF-dev considers an N sac instead of B sac and finds nothing better than a draw. It's interesting that it finds Bxg6 on the very first ply but immediately rejects it

Code: Select all

  40	  0.00 	723.3M	14:37.16	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 39	  0.00 	539.2M	10:47.12	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 38	  0.00 	411.4M	7:49.51	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 37	  0.00 	348.2M	6:30.79	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 36	  0.00 	260.7M	4:52.26	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 35	  0.00 	220.9M	4:09.16	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 34	  0.00 	159.4M	3:00.63	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 33	  0.00 	129.3M	2:24.57	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 32	  0.00 	105.5M	1:57.98	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 31	  0.00 	96.5M  	1:47.70	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 Kf3 Qb4 Bf5 Re8 Qg7+ Re7 Bxe6+ Kxe6 Rc6+ Bxc6 Qf6+ Kd7 Qxc6+ Kd8 Qa8+ Kc7 Qa7+ Kc8 Qa8+ Kd7 Qc6+
 30	  0.00 	69.1M  	1:17.59	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 f5 exf5 Bxf5+ Kd8 Ke2 Qf7 Bg6 Qe7 Bf5 Qf7
 29	  0.00 	61.1M  	1:08.41	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 f5 exf5 Bxf5+ Kd8 Ke2 Qf7 Bg6 Qe7 Bf5 Qf7
 28	  0.00 	49.5M  	0:55.47	Nxe6 fxe6 Bxg6 Qf8 h5 Ba3 f5 exf5 Bxf5+ Kd8 Ke2 Qf7 Bg6 Qe7 Bf5 Qf7
 28	+0.13?	45.2M  	0:50.72	Nxe6 fxe6?
 28	+0.42?	39.5M  	0:44.19	Nxe6 fxe6?
 28	+0.64?	35.6M  	0:39.84	Nxe6 fxe6?
 28	+0.80?	34.0M  	0:38.03	Nxe6 fxe6?
 28	+0.91?	31.6M  	0:35.39	Nxe6 fxe6?
 28	+1.06!	31.1M  	0:34.81	Nxe6!
 27	+0.98!	24.1M  	0:27.16	Nxe6!
 27	+0.77!	21.1M  	0:23.87	Nxe6!
 27	+0.61!	20.0M  	0:22.76	Nxe6!
 27	+0.50!	19.4M  	0:22.05	Nxe6!
 27	+0.42!	18.5M  	0:21.06	Nxe6!
 26	+0.35 	17.1M  	0:19.59	Nxe6 fxe6 Bxg6 Qf8 h5 Rg8 Ke2 Qg7 f5 Qh6 Kf3 Rc8 fxe6+ Kd8 Qf4 Qxf4+ Kxf4 Rc4 Rxc4 dxc4 h6 Bf8 Kg5 Ke7 Bf5 c3 Kg6 Bxh6 Kxh6 Bd5 Bc2 Kxe6 Kg5 Bc4 Kf4 Kd5
 26	+0.72!	16.9M  	0:19.27	Nxe6!
 26	+0.56!	16.2M  	0:18.54	Nxe6!
 26	+0.45!	15.4M  	0:17.68	Nxe6!
 26	+0.37!	14.7M  	0:16.89	Nxe6!
 25	+0.30 	12.3M  	0:14.25	Nxe6 fxe6 Bxg6 Qf8 h5 Rg8 Ke2 Qg7 f5 Qh6 Kf3 Rc8 fxe6+ Kd8 Qf4 Qg7 Qf7 Qxf7+ exf7 Rxc2 Bxc2 Bf8 Kg4 Ke7 Kg5 Kxf7 Bg6+ Ke7 h6 Bc8 Bh5 Be6 h7 Bg7 Kg6
 25	+0.35?	11.6M  	0:13.58	Nxe6 fxe6?
 25	+0.60!	10.8M  	0:12.66	Nxe6!
 25	+0.51?	10.6M  	0:12.39	Nxe6 fxe6?
 25	+0.62?	10.4M  	0:12.22	Nxe6 fxe6?
 25	+0.70?	10.3M  	0:12.05	Nxe6 fxe6?
 24	+0.77 	9.67M  	0:11.37	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Be7 Ke2 Rg8 f5 Qh6 Kf3 Rf8 f6 Bb4 Bd3 Kd8 Qg6 Bd2 Qxh6 Bxh6 Rg2 Rh8 Rg6 Ke8 Rg7 Bxg7 fxg7 Rg8 h6 Kf7 h7 Kxg7 hxg8=Q+ Kxg8 Kf4
 24	+0.72!	7.58M  	0:09.01	Nxe6!
 24	+0.51!	6.74M  	0:08.09	Nxe6!
 24	+0.35!	5.94M  	0:07.22	Nxe6!
 24	+0.24!	5.77M  	0:07.06	Nxe6!
 24	+0.16!	5.72M  	0:07.00	Nxe6!
 23	+0.09 	5.49M  	0:06.78	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rc8 fxe6+ Kd8 Rxc8+ Bxc8 Bf5 Kc7 e7 Bxf5 e8=Q Bxg4+ Kxg4 Bd2 Qe7+ Kc8 Qf6 Kd7 Qf7+ Kd8 Kf5 Be3 Qf6+ Kd7 Qxh6 Bxh6
 23	+0.15!	5.20M  	0:06.39	Nxe6!
 23	+0.07!	5.04M  	0:06.22	Nxe6!
 22	  0.00 	4.73M  	0:05.88	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rc8 fxe6+ Kd8 Rxc8+ Bxc8 Bf5 Kc7 e7 Bxf5 e8=Q Bxg4+ Kxg4 Bd2 Qe7+ Kc8 Qf6 Kd7 e6+ Kd6 Qe5+ Ke7 Qc7+ Kxe6 Qxb6+ Kd7 Qb7+ Ke8 Qb8+ Ke7 Qe5+ Kf8 Qb8+ Kg7 Qb7+ Kf8 Qb8+
 21	  0.00 	4.40M  	0:05.47	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rc8 fxe6+ Kd8 Rxc8+ Bxc8 Bf5 Kc7 e7 Bxf5 e8=Q Bxg4+ Kxg4 Bd2 Qe7+ Kc8 Qf6 Kd7 e6+ Kd6 Qe5+ Ke7 Qc7+ Kxe6 Qxb6+ Kd7 Qb7+ Ke8 Qb8+ Ke7 Qe5+ Kf8 Qb8+ Kg7 Qb7+ Kf8 Qb8+
 21	+0.12?	4.20M  	0:05.22	Nxe6 fxe6?
 21	+0.24?	3.96M  	0:04.94	Nxe6 fxe6?
 21	+0.31?	3.81M  	0:04.76	Nxe6 fxe6?
 20	+0.39 	3.51M  	0:04.39	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rc8 fxe6+ Kd8 Rxc8+ Bxc8 Qh4+ Be7 Qf4 Qxf4+ Kxf4 Bf8 Kf5 Ke7 Kg5 Bxe6 h6 Bh3 Bh5 Ke6 h7 Bg7 Kg6
 20	+0.52!	3.16M  	0:03.93	Nxe6!
 20	+0.27?	2.99M  	0:03.71	Nxe6 fxe6?
 20	+0.61!	2.67M  	0:03.33	Nxe6!
 20	+0.50!	2.55M  	0:03.13	Nxe6!
 19	+0.35 	2.04M  	0:02.55	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rf8 Qf4 Qxf4+ Kxf4 Rc8 Rg2 Rh8 fxe6+ Kxe6 Bf5+ Kf7 Rg6 Bd2+ Kg4 Bc8 Bxc8 Rxc8 Rxb6 Rc4 Kf5 Rxa4 Rb7+ Ke8
 18	  0.00 	1.69M  	0:02.14	Nxe6 fxe6 Bxg6 Qf8 h5 Bb4 Kf3 Qh6 f5 Rf8 Qf4 Qxf4+ Kxf4 Rc8 Rxc8 Bd2+ Kf3 Bxc8 f6 Bb7 Ke2 Bf4 Kf3 Bg5 Ke2 Bh6
 17	  0.00 	1.27M  	0:01.64	Rh2 Rh6 Rh1 Ba3 f5 gxf5 Bxf5 exf5 Qxf5+ Kc7 Nxf7 Re6 Ng5 Rh6 Nf7
 16	  0.00 	554153	0:00.75	Rh2 Ba3 Rc2 Be7 Rh2
 15	  0.00 	506281	0:00.67	Rh2 Ba3 Rc2 Be7 Rh2
 14	  0.00 	473867	0:00.64	Rh2 Ba3 Rc2 Be7 Rh2
 13	  0.00 	441835	0:00.59	Rh2 Bd8 Qe2 Be7 Qg4
 12	  0.00 	98223  	0:00.16	Rc3 Qf8 Rc1 Bd8 Ke2 Qa3 Rc2 Qf8 Rc1
 11	  0.00 	82113  	0:00.13	Rc3 Qf8 Rc1 Bd8 Ke2 Qa3 Rc2 Qf8 Rc1
 10	  0.00 	78428  	0:00.13	Rc3 Qf8 Rc1 Bd8 Ke2 Qa3 Rc2 Qf8 Rc1
  9	  0.00 	23103  	0:00.03	Kf2 Qf8 Nf3 Ke8 Ng5 Kd7
  8	+0.12 	11936  	0:00.02	f5 gxf5 Bxf5 Qg8 Bd3 Qf8 Qf4 Ke8
  7	+0.35 	10001  	0:00.02	f5 Bxg5+ Qxg5 gxf5 Kf4 Qf8 Kg3
  6	+0.87 	2378    	0:00.01	Bf1 Bxg5 fxg5 Qe7 Bd3 Rc8
  5	+1.92 	447      	0:00.00	Bf1 Bd8 Bh3 Bxg5 Qxg5
  4	+1.92 	367      	0:00.00	Bf1 Bd8 Bh3 Bxg5 Qxg5
  3	+2.15 	201      	0:00.00	Nf3 g5 Nxg5
  2	+2.15 	152      	0:00.00	Nf3 g5 Nxg5
  1	+1.40 	55        	0:00.00	Bxg6
  0	#


User avatar
Rebel
Posts: 4174
Joined: Thu Aug 18, 2011 10:04 am

Re: MCTS-NN vs alpha-beta

Post by Rebel » Thu Dec 07, 2017 10:16 am

Can't believe it without a press release from Google.

corres
Posts: 532
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Historic Milestone: AlphaZero

Post by corres » Thu Dec 07, 2017 10:54 am

[quote="maac"]

Take note that SF was at 64 cores!

[/quote]

Another VERY IMPORTANT notes:
1, SF used only 1 GB (!!) hash table
2, Alpha Zero did not start from zero knowledge about chess
because it was feeded a lot of human games at start up.
This is the explanation why Alpha Zero plays openings so human like.
I think it would be more correct if Stockfish would get 64 GB hash
and a good human opening book like Fritz Power Book.

duncan
Posts: 7346
Joined: Mon Jul 07, 2008 8:50 pm

Re: Historic Milestone: AlphaZero

Post by duncan » Thu Dec 07, 2017 11:36 am

JJJ wrote:ALphaZero did win Stockfish in 100 game scored 64/100 that is 98 elo stronger than Stockfish TCEC 2016. Not bad ! I d like to see it more trained to see how far from perfect Stockfish really is !
and it did this after just 4 hours training. ? not sure why they did not train for at least a week to get an even better result.

duncan
Posts: 7346
Joined: Mon Jul 07, 2008 8:50 pm

Re: Historic Milestone: AlphaZero

Post by duncan » Thu Dec 07, 2017 11:39 am

maac wrote:https://arxiv.org/pdf/1712.01815.pdf

To say that i'm impressed is a understatement. Shocking.
Take note that SF was at 64 cores!
25 wins for white and only 3 wins for black. is such a big difference what you would expect ?

Werewolf
Posts: 994
Joined: Thu Sep 18, 2008 8:24 pm

Re: Historic Milestone: AlphaZero

Post by Werewolf » Thu Dec 07, 2017 11:50 am

corres wrote:
maac wrote:
Take note that SF was at 64 cores!
Another VERY IMPORTANT notes:
1, SF used only 1 GB (!!) hash table
2, Alpha Zero did not start from zero knowledge about chess
because it was feeded a lot of human games at start up.
This is the explanation why Alpha Zero plays openings so human like.
I think it would be more correct if Stockfish would get 64 GB hash
and a good human opening book like Fritz Power Book.
THIS.

1 GB hash allows them to say SF nps are really high while hiding that they’ve weakened its search.

I would like to see a rematch at a better time control, much more hash, all tablebases, and a tournament quality opening book and the latest asmfish.

I bet it’d be much closer then.

Lyudmil Tsvetkov
Posts: 6036
Joined: Tue Jun 12, 2012 10:41 am

Re: MCTS-NN vs alpha-beta

Post by Lyudmil Tsvetkov » Thu Dec 07, 2017 12:43 pm

MikeGL wrote:
kranium wrote:
Lyudmil Tsvetkov wrote:It is not at all clear to me where were books used and where not.
I'm sure opening books were not used...
In the early self-play games things like 1.a3, 1.a4, etc. were probably tried by AlphaZero...
eventually it learned that 1. e4 or 1. d4 had the highest success rates.
Books or no books, I think AlphaZero would still demolish SF8.
Just look at this game 9, it was a decent French Defence by SF8, but it was dismantled with
amazing tactical and strategic shots by AlphaZero which seems to be beyond the reach of alpha-beta engines.

[pgn]
[Event "?"]
[Site "?"]
[Date "2017.12.06"]
[Round "9"]
[White "AlphaZero"]
[Black "Stockfish"]
[Result "1-0"]
[TimeControl "40/1260:300"]
[Termination "normal"]
[PlyCount "103"]
[WhiteType "human"]
[BlackType "human"]

1. d4 e6 2. e4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 cxd4 7. Nb5 Bb4+ 8.
Bd2 Bc5 9. b4 Be7 10. Nbxd4 Nc6 11. c3 a5 12. b5 Nxd4 13. cxd4 Nb6 14. a4
Nc4 15. Bd3 Nxd2 16. Kxd2 Bd7 17. Ke3 b6 18. g4 h5 19. Qg1 hxg4 20. Qxg4
Bf8 21. h4 Qe7 22. Rhc1 g6 23. Rc2 Kd8 24. Rac1 Qe8 25. Rc7 Rc8 26. Rxc8+
Bxc8 27. Rc6 Bb7 28. Rc2 Kd7 29. Ng5 Be7 30. Bxg6 Bxg5 31. Qxg5 fxg6 32. f5
Rg8 33. Qh6 Qf7 34. f6 Kd8 35. Kd2 Kd7 36. Rc1 Kd8 37. Qe3 Qf8 38. Qc3 Qb4
39. Qxb4 axb4 40. Rg1 b3 41. Kc3 Bc8 42. Kxb3 Bd7 43. Kb4 Be8 44. Ra1 Kc7
45. a5 Bd7 46. axb6+ Kxb6 47. Ra6+ Kb7 48. Kc5 Rd8 49. Ra2 Rc8+ 50. Kd6 Be8
51. Ke7 g5 52. hxg5 1-0
[/pgn]

not sure if 18.g4!, 30.Bxg6! and other would be found by current engines.

[d]r2qk2r/3bbppp/1p2p3/pP1pP3/P2P1P2/3BKN2/6PP/R2Q3R w kq - 0 18
After 17...b6 of black, can some engine consider 18.g4! in this position?



xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
[d]4q2r/1b1kbp2/1p2p1p1/pP1pP1N1/P2P1PQP/3BK3/2R5/8 w - - 6 30
After 29...Be7, can current engines consider 30.Bxg6! here?


Would be nice if we can try to feed some difficult epd positions into AlphaZero,
to estimate its ELO strength.
I am considering g4 and Bg6, especially g4 was my first choice in under a second.
People don't believe frequently advanced long chains are stronger than a pice, but see what happens here...

I guess the main fault is thay are testing at 1 minute. Long chains have failed in SF at least 5 times or so, and still they are an extremely valid concept.

Books are important, as I have been correctly claiming for a very long time, the French is dangerous or even lost, but people have hard ears.
The game was lost much earlier, already in the opening.

User avatar
MikeGL
Posts: 700
Joined: Thu Sep 01, 2011 12:49 pm

Re: MCTS-NN vs alpha-beta

Post by MikeGL » Thu Dec 07, 2017 12:49 pm

Lyudmil Tsvetkov wrote:
MikeGL wrote:
kranium wrote:
Lyudmil Tsvetkov wrote:It is not at all clear to me where were books used and where not.
I'm sure opening books were not used...
In the early self-play games things like 1.a3, 1.a4, etc. were probably tried by AlphaZero...
eventually it learned that 1. e4 or 1. d4 had the highest success rates.
Books or no books, I think AlphaZero would still demolish SF8.
Just look at this game 9, it was a decent French Defence by SF8, but it was dismantled with
amazing tactical and strategic shots by AlphaZero which seems to be beyond the reach of alpha-beta engines.

[pgn]
[Event "?"]
[Site "?"]
[Date "2017.12.06"]
[Round "9"]
[White "AlphaZero"]
[Black "Stockfish"]
[Result "1-0"]
[TimeControl "40/1260:300"]
[Termination "normal"]
[PlyCount "103"]
[WhiteType "human"]
[BlackType "human"]

1. d4 e6 2. e4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 cxd4 7. Nb5 Bb4+ 8.
Bd2 Bc5 9. b4 Be7 10. Nbxd4 Nc6 11. c3 a5 12. b5 Nxd4 13. cxd4 Nb6 14. a4
Nc4 15. Bd3 Nxd2 16. Kxd2 Bd7 17. Ke3 b6 18. g4 h5 19. Qg1 hxg4 20. Qxg4
Bf8 21. h4 Qe7 22. Rhc1 g6 23. Rc2 Kd8 24. Rac1 Qe8 25. Rc7 Rc8 26. Rxc8+
Bxc8 27. Rc6 Bb7 28. Rc2 Kd7 29. Ng5 Be7 30. Bxg6 Bxg5 31. Qxg5 fxg6 32. f5
Rg8 33. Qh6 Qf7 34. f6 Kd8 35. Kd2 Kd7 36. Rc1 Kd8 37. Qe3 Qf8 38. Qc3 Qb4
39. Qxb4 axb4 40. Rg1 b3 41. Kc3 Bc8 42. Kxb3 Bd7 43. Kb4 Be8 44. Ra1 Kc7
45. a5 Bd7 46. axb6+ Kxb6 47. Ra6+ Kb7 48. Kc5 Rd8 49. Ra2 Rc8+ 50. Kd6 Be8
51. Ke7 g5 52. hxg5 1-0
[/pgn]

not sure if 18.g4!, 30.Bxg6! and other would be found by current engines.

[d]r2qk2r/3bbppp/1p2p3/pP1pP3/P2P1P2/3BKN2/6PP/R2Q3R w kq - 0 18
After 17...b6 of black, can some engine consider 18.g4! in this position?



xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
[d]4q2r/1b1kbp2/1p2p1p1/pP1pP1N1/P2P1PQP/3BK3/2R5/8 w - - 6 30
After 29...Be7, can current engines consider 30.Bxg6! here?


Would be nice if we can try to feed some difficult epd positions into AlphaZero,
to estimate its ELO strength.
I am considering g4 and Bg6, especially g4 was my first choice in under a second.
People don't believe frequently advanced long chains are stronger than a pice, but see what happens here...

I guess the main fault is thay are testing at 1 minute. Long chains have failed in SF at least 5 times or so, and still they are an extremely valid concept.

Books are important, as I have been correctly claiming for a very long time, the French is dangerous or even lost, but people have hard ears.
The game was lost much earlier, already in the opening.
LOL there goes your rediculous claims again. You haven't claimed French is lost. Lets be
honest please, you only claimed French WINAWER is lost, nothing else.
You claimed all other variations of French are playable but not the WINAWER line.

Winawer discussion was a thread on its own, I remember clearly.
Last edited by MikeGL on Thu Dec 07, 2017 12:51 pm, edited 1 time in total.

Post Reply