Author: Heiner Marxen
Date: 14:56:44 06/18/00
Go up one level in this thread
On June 18, 2000 at 07:14:20, Gian-Carlo Pascutto wrote: > >Hi all > >At the bottom of this post is a test position. The idea is not to check which >programs find the correct move (should be easy), but when the program first >spots it, with which score, when it fails high, how long it takes for the >fail-high to resolve and what score it gets back. > >The idea is to see how well the programs handle this kind of positions where >the extensions suddenly can get wild, just when the alpha-beta window opens. >Some can handle this very well, other have the search totally blow up... > >I've done a bit of testing already, and the results vary rather widely...so >I'd like to get as much data as possible on this one. If you are a programmer >yourself, _please_ indicate what extensions your program uses, or other >relevant info about your search. I'd like to get an idea of the do's/dont's >when extending. > >Also, if anyone has other positions like this (preferrably even harder/worse), >please share them. > >[D] rr4k1/pp2pp1p/q2p2p1/2pP3n/2P5/2BQP2R/PP3P2/2K4R w - - 0 19 > >PS. Is this a forced mate? If so, how many moves exactly? According to Chest this is a mate in 8 (570 seconds on a K7/600, 350 MB hash). 1.Rxh5 is the unique solution. Here is the timing over the depth: # 1 0.00 0.87 1- 0 # 2 0.00 1.00 1- 0 # 3 0.01 0.98 50- 0 # 4 0.09 1.13 847- 0 # 5 0.58 1.41 5750- 0 # 6 3.23 1.69 29474- 0 # 7 33.24 1.94 259368- 0 # 8 569.54 2.30 3736465- 4341 There is a significant increase in the last level, but that is not uncommon for the solution depth. But when Chest prints the (more or less) complete solution tree, I find it somewhat special: Even after some compression by "any" and "else" notation (for the defender) the dual free tree has ca 1300 lines (terminal nodes). With duals (alternate solution moves) this explodes to ca 77000, even without mate move duals. By skimming through it I find that many of these lines are full width variations. That may explain an aspect of the problem you encounter. Heiner >-- >GCP
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.