Author: Uri Blass
Date: 21:00:21 09/11/01
Go up one level in this thread
On September 11, 2001 at 22:44:00, Robert Hyatt wrote: >On September 11, 2001 at 12:18:52, Uri Blass wrote: > >>On September 11, 2001 at 10:26:21, Robert Hyatt wrote: >> >>>On September 11, 2001 at 02:17:09, Uri Blass wrote: >>> >>>>On September 10, 2001 at 21:27:02, Dave Gomboc wrote: >>>> >>>>>On September 10, 2001 at 05:56:21, Uri Blass wrote: >>>>> >>>>>>On September 10, 2001 at 05:02:43, Gian-Carlo Pascutto wrote: >>>>>> >>>>>>>On September 10, 2001 at 00:48:34, Dave Gomboc wrote: >>>>>>> >>>>>>>>>>>>>>32.Bg5 Bg7 33.Bf4 c4 34.Re2 Bd7 35.Rd1 Bb5 36.Ree1 Bxa4 37.bxa4 Rb2 >>>>>>>>>>>>>>38.Rd2 Rxa2 39.Bxf5 >>>>>>> >>>>>>>>>>Black has better earlier on: >>>>>>>>>> >>>>>>>>>>Instead of 36. .. Bxa4 first 36. .. Nd3! >>>>>>>> >>>>>>>>It looks like White also has better earlier on: >>>>>>>> >>>>>>>>36.Bxe5! Nxe5 >>>>>>> >>>>>>>Bxe5 captures the knight, so Nxe5 is obviously impossible. I think >>>>>>>you are looking at the wrong line... >>>>>>> >>>>>>>36.Bxe5 Bxe5 is a straight win for black >>>>>> >>>>>>Dave meant 36.Bxe5 Bxe5 and it is not clear >>>>>> >>>>>>Here is again the relevant analysis(I added * to singular moves of black that is >>>>>>before move 39 and I am going to look at this position more today because I have >>>>>>other things to do) >>>>>> >>>>>>Thanks to dave Gomboc for doing the job. >>>>>>I did not look at most of the analysis for 37...Bxh2+ >>>>>> >>>>>>36.Bxe5! Bxe5* 37.Nc5! Bxh2+ 38.Kxh2 dxc5* 39.a4 Bc6* 40.Rd6 cxb3 41.Bd3 b2 >>>>>>42.Re1 b1Q 43.Rxb1 Rxb1 44.Bxb1 is unclear, while after 40.Re6 Kf7 41.Rde1 cxb3 >>>>>>42.Bd3 Bb5 43.axb5 Rxe6 44.Bc4 Rbb6 45.Bxb3 White should be able to hold the >>>>>>rook ending. Meanwhile, 40.Re6 Kf7 41.Rde1 Rxd7 42.b4 cxb4 43.cxb4 Rb7 44.Bd1 >>>>>>Rxb4 45.Bf3 Bxf3 46.Rxa6 Rb2 is wild!, I'll let someone with a faster machine >>>>>>handle that variation (there may be good deviations for either side along the >>>>>>way too). If something refutes this line, it's probably this. >>>>>> >>>>>>This leaves 36.Bxe5! Bxe5* 37.Nc5! dxc5 38.Rxe5 Rxa2* 39.bxc4 Bxc4 40.Rxc5 which >>>>>>looks like White should be able to grovel a draw here too. >>>>>> >>>>>>39...Bc6 40.Rxc5 Bxg2 may be an improvement for black and I did not check this >>>>>>line >>>>>> >>>>>>Uri >>>>> >>>>>I think more of the moves are singular than you are giving credit for. For >>>>>instance, 38.Kxh2 is an obvious case (though perhaps you were counting that as a >>>>>normal recapture-extension). >>>> >>>>I count only black moves and I did not count moves after move 39. >>>> >>>>Uri >>> >>> >>>If we are talking about SE, then both sides can be extended. >> >>I counted only black moves for the analysis. >>It was not to discuss about their singular extensions. >> >>There was a claim that white can draw the position and singular lines for black >>are lines that you do not need to search for improvements for black. >> >> You have to take >>>their output, ie 11(6) and use that "11 plies" as a starting point. >> >>11(6) may be relevant for the deeper blue games but not for this game because it >>was Deep thought game and I understood that Deep thought searched only 10-12 >>plies+singular extensions >> >> Each >>>singular move adds 1 ply to that. Note that a move can be both singular and >>>out of check, and it gets extended by 2 plies. Their only extension constraint >>>was that two consecutive plies could not extend more than 2 plies total. IE if >>>you extend twice at ply N, then you could not extend by 1 on both the previous >>>and next moves (before and after that 2-ply extension ply). >>> >>>That kind of search can probe _very_ deeply in the right lines since SE was >>>not their _only_ extension. It was generally "in addition to" everything else >>>we all normally do. >> >>With all these extensions their search depth should be enough to see draw by >>repetition or material equality against kasparov at game 2 of the match. > > >I don't see how. That draw is some 60 plies deep. I don't think any program >can search to a depth of 60 plies, following checks like that.The size of the >tree is enormous, even at 200M nodes per second. 1)The size of the tree is not enormous. If it was so big humans had no chance to prove the draw or equal material by a tree. I proved it by a tree when programs could see draw evaluation at every leaf of the tree after a short search. If you count the number of nodes in all the relevant small trees of short search you can see that the tree is not enormous. programs of today do not use singular extensions and my point is that with singular extensions the size of the tree is not enormous so saying that programs of today cannot see the draw is not a convincing claim. 2)It is also interesting that you did not claim that the size of the tree is enormous about the c5 move. This is exactly the claim of people who believe that deep thought could not see +2 score against Cray blitz. For some reason there is no limit for the number of plies that Deep thought could see by singular extensions and even if the +2 evaluation happens only 60 plies after c5 Gian-Carlo Pascutto is going to use it in the discussion if singular extension could help deep thought to see +2.xx evaluation. Gian-Carlo Pascutto used lines of more than 40 plies after c5 and I say 60 plies because the evaluation in the leaf was not static evaluation but evaluation that is based on search. Deep thought was also clearly weaker than Deeper blue so we can assume that if Deeper blue could not see 60 plies forward then Deep thought could not see even 50 plies forward even with singular extensions. Uri
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.