Author: Heiner Marxen
Date: 16:13:45 07/12/01
Go up one level in this thread
On July 12, 2001 at 14:07:59, leonid wrote: >Hi! > >If you like, you can solve this mate: > >[D]krqn4/rqP3qq/n1QQ4/bbNNBBqq/q1QQ4/Q1P3qq/RQ6/KRN5 w - - > >I found very intriguing few anomalies about this position. NPS (node/per/second) >rate was very high while braching factor was unusually bad. This is true for >brute force search. NPS was 912k (8 plys search) when it could be expected as >low as 45k. When I went to try this position on Rebel 10, I found that Rebel's >NPS is, for the first time, lower that mine. Genius 4 time, for 8 plys search, >was almost the same as mine. Only from 10 plys my program started gaining some >modest distance. I never could see nothing like this before while verifying my >"normal" positions. > >Please indicate your result, > >Thanks, >Leonid. Hmmm, for Chest this one is a bit different than your "normal stuff" in so far as it needs more time to reach a certain depth. On a K7/600 with 350 MB hash Chest needs already 80 minutes just to prove that there is no mate in 7. And with an effective branching factor above 10 Chest is not going to go much deeper. Some timing info: depth time EBF[T] EBF[N] # 1 0.00s 0kN 0.94 1- 0 # 2 0.01s 0kN [146.50] 1.00 1- 0 # 3 0.18s [ 18.00] 10kN [ 33.82] 0.99 257- 0 # 4 2.54s [ 14.11] 154kN [ 15.57] 1.24 5568- 0 # 5 31.91s [ 12.56] 1894kN [ 12.27] 1.58 81370- 0 # 6 427.53s [ 13.40] 25434kN [ 13.43] 1.94 1241355- 2 # 7 4837.53s [ 11.32] 272444kN [ 10.71] 2.60 15523159- 6789678 272444kN / 4837.53s = 56.3 kNPS, which is pretty normal for Chest. Also, the branching factors, as measured by time (EFB[T]) and by nodes (EBF[N]) appear to be quite normal for such a board, where black does not have any counter checks. So, I'm not sure what makes this position special for your program, since for Chest it does not appear to be exceptional. Since I do some other heavy chess problems now, I have stopped further analysis of this position, sorry. Cheers, Heiner
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.