Author: Uri Blass
Date: 01:57:40 06/04/01
Go up one level in this thread
On June 04, 2001 at 04:35:19, Uri Blass wrote: >On June 04, 2001 at 03:36:25, Rudolf Huber wrote: > >> >>A couple of days ago I said (in a discussion about mtd(f)) that my chess >>program can search the starting position to 30 ply over night when the eval >>evaluates material only: >> >> >>SOS no eval >> 1.00 0:00 0.00 1.h3 (2) >> 2.00 0:00 0.00 1.h3 (45) >> 3.00 0:00 0.00 1.h3 h6 2.h4 (138) >> 4.00 0:00 0.00 1.h3 h6 2.h4 (263) >> 5.00 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 (435) >> 6.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 (1.860) >> 7.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 (1.942) >> 8.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 (2.024) >> 9.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 (5.086) >> 10.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 (16.891) >> 11.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (41.424) >> 12.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (94.654) >>463.9 >> 13.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (132.293) >>497.3 >> 14.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (245.124) >>505.4 >> 15.01 0:00 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (359.203) >>533.7 >> 16.01 0:01 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (714.937) >>593.8 >> 17.01 0:02 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (1.212.436) >>570.2 >> 18.01 0:02 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (1.729.490) >>585.4 >> 19.01 0:05 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (3.025.316) >>605.0 >> 20.01 0:08 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (4.700.415) >>587.5 >> 21.01 0:15 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (9.323.868) >>621.5 >> 22.01 0:29 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (16.881.247) >>582.1 >> 23.01 0:48 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (28.659.953) >>597.0 >> 24.01 1:36 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 (56.472.536) >>588.2 >> 25.01 3:37 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(127.308.632) 586.6 >> 26.01 7:41 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(275.114.597) 596.7 >> 27.01 17:31 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(622.179.104) 591.9 >> 28.01 45:16 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(1.602.666.536) 590.0 >> 29.01 80:46 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(-1452508008) 586.5 >> 30.01 161:28 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(1.397.948.590) 587.6 >> 31.01 513:07 0.00 1.h3 h6 2.h4 h5 3.g3 g6 4.Rh3 Rh6 5.Rh1 Rh8 >>(471.878.568) 573.3 >> >> >>Actually 30 ply was reached after 2:41 hours. >> >>SW: no search extensions, normal nullmove pruning, normal quiescence >> search >>HW: 2xCeleron 466 MHz, 500 MByte Hash >> >> >>Can anyone do better? >Some questions: > >1)Is normal null move pruning recursive or not recursive null move pruning? >In other words is it possible that your program will miss something that is 15 >plies when it searches 30 plies not because of zunzwang. > >2)What is the result without null move pruning? >It will be interesting to compare your results with Deeper blue depthes in the >positions from kasparov-deeper blue games. > >3)I see that you get depth 1 with only 2 nodes > >I guess that the tree include only the following 2 positions >the beginning position and the position after h3. > >Am I right? > >I guess that you use the qsearch at the last ply because you do not need to >generate all the moves in the last ply but only captures when your evaluation is >only material evaluation. > >4)How do you get 45 nodes at depth 2? > >I see that the tree is very small and even at depth 3 you get only 138 nodes. >Can you post the full tree for the first depthes. > >I think it may be productive for people to understand exactly what is the >algorithm that your program use. > >I prefer to look at things by examples and not by formal definitions that may be >more complicated. > >Uri I can add that this is my way of looking at things and other people may have other opinion. I am not against formal defintions but I prefer to see first examples and from the examples I can build formal definitions to explain the examples I try to use the examples pf number of nodes in order to understand your algorithm and I think that I understood a rule that explain the numbers for depthes 1,2 I guess 2 nodes at depth 1 is because you first use the qsearch to see that there is no captures and after it generates one node and evaluates it(it is obvious that all the other will be the same) If we use the same principle at depth 2 I get: 1)21 obvious nodes that are root position and positions after one ply. 2)There are no captures after the main line h3 so it is enough to generate one move to get 1.h3 h6 and get evaluation 0.00 3)I need to search for a killer move for every other move that is not 1.h3 so I get 19 lines 1.h4 h6,1.g3 h6,1.g4 h6.... 4)There are captures after some of the lines(I mean d3 h6 and d4 h6 so you need to search 1.d3 h6 2.Bxh6 Rxh6 1.d4 h6 Bxh6 Rxh6 so you get more 4 nodes. Total number of nodes seem to be 21+1+19+4=45 I needed a time to see 4 and this is the reason that I did not understand 45 in a short time. Uri
This page took 0 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.