Author: Ulrich Tuerke
Date: 01:57:38 05/23/03
Go up one level in this thread
On May 22, 2003 at 12:11:33, José Carlos wrote: >On May 22, 2003 at 11:42:45, Ulrich Tuerke wrote: > >>On May 22, 2003 at 11:16:18, David H. McClain wrote: >> >>>On May 22, 2003 at 08:41:56, Ulrich Tuerke wrote: >>> >>>>On May 22, 2003 at 06:08:47, José Carlos wrote: >>>> >>>>>On May 22, 2003 at 01:01:38, Robin Smith wrote: >>>>> >>>>>>Often programs will underpromote a pawn in situations where they think capturing >>>>>>the (under)promoted piece is still the best reply. Can anyone give an example >>>>>>from an actual game where this underspromotion cost the computer a 1/2 point? >>>>>> >>>>>>Robin >>>>> >>>>> I can't recall any game with this problem, but it's really easy to fix. I >>>>>simply order the promotions so that promoting to queen gets searched first. >>>> >>>>Isn't everybody doing this anyway ? >>>>Uli >>>> >>>>> The >>>>>all the other underpromotions will return alpha and thus ignored. >>>>> >>>>> José C. >>> >>>Jose, >>> >>>No, everyone is not doing this. At least I am not. Admittedly I have played >>>only one game where an underpromotion was an absolute necessity to regain tempo >>>and obtain a "check" to prevent a checkmate to the other side. This was to a >>>knight promotion rather than a Queen. And the game was eventually won with the >>>knight promotion. I'm sure others have witnessed this on occasion also. >>> >>>Don't ask me to find and post this game. I don't know where or if I have it but >>>I simply regarded it as "the program knows best" and let it go at that. The >>>program was correct.......... >> >>I don't think that we are talking about omitting under-promotions but about >>move-ordering solely. >> >>When generating moves in full search, I sort promotions like >> >>1. queen >>2. knight >>3. rook >>4. bishop >> >>That's because rook or bishop promotions will only very rarely generate a cut in >>a case where the queen doesn't. Usually, the queen will do anyway. >> >>However, the knight may be a serious alternative to a queen promotion. >>Therefore, I prefer it to rook and bishop. >> >>This way, you can't miss a win by knight promotion. >> >>In quies search, I omit (3) and (4). >> >>Uli > > I recently implemented ignoring underpomotions in quies, because I read it in >Ed's page (haven't figured it out myself before) and though about, as you do, >try knight promotions (Ed doesn't). The problem is finding positions to test it. > I think (haven't tested) that the need of underpromotions to knight in quies >must be so rare in real life that Ed's approach should be better. Have you >tested with and without? No, I haven't even read Ed's suggestions. I have not been busy with chess programming for some time now. I guess, it's rather a tiny effect: perhaps you miss to solve one position out of thousands, and you save 1 per-mille of nodes. -:) Uli > > José C.
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.