Author: Uri Blass
Date: 00:43:14 03/05/05
Go up one level in this thread
On March 05, 2005 at 02:00:21, Reinhard Scharnagl wrote: >On March 04, 2005 at 17:33:44, Russell Reagan wrote: > >>On March 04, 2005 at 04:36:19, Reinhard Scharnagl wrote: > >Hi Russel, > >>>We handle here valid chess positions of the traditional chess game >>>and calculate codelengths for a special subset of RELEVANT positions >>>matching ONE simple restriction (without relevance for practice): >>> >>> "For each performed promotion there also must have been captures >>> during the game: one pawn or half an officer." > >>I think this assumption is not correct, or maybe I misunderstand it. It is >>possible to promote two white pawns for each black pawn captured. To promote all >>8 white pawns only requires 4 captured black pieces. Then you can promote all 12 >>pawns. See my (legal) game below which demonstrates this. > >this is a practical restriction, not an assumption. > >What I claim is, that really played games will follow that assumption. > >But because of a lot to be saved bits, even numerous positions with some such >unlikely promotions would fit into that length scheme. > >Reinhard. You can probably make more practical assumptions: 1)no more than x promotions for one side.(need to analyze games to find x but I guess that x<=4 for all interesting positions because if a sude can win by promoting 5 queens the same side can win earlier by mate) 2)no more than y underpromotions for one side(I guess y<=3). 3)no more than z pawn captures(even without that assumption there are many illegal pawn structures like white pawns at a2 b2 a3 or white pawns at a2 a3 a4 when black has at least 14 pieces on the voard). 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.