Author: James Robertson
Date: 18:46:48 10/29/98
Go up one level in this thread
On October 29, 1998 at 19:01:24, James Long wrote: >Hello all, > >I've been playing a bit lately with extensions. >One issue I've been wrestling with is when to say "enough." > >I limit the number of extensions that can be done >in one node to two ply. I'd be interested to hear >if anyone has played with limiting the extensions >throughout an entire branch. > >The two problems below are from WAC. The first run >is from Tristram v3.37, the second from what will soon >be v4.0. In *this* case, the extensions appear to >be a win. Actually, they appear to be a win >throughout the suite, but I've seen a few instances >on ICC that the program appeared to take too long >to get through an iteration. Even though the second >run came up with a mate (an easy one), I'm concerned >with the fact that the first iteration went nine moves deep. > >Similiar experiences? > > >position 009 >3q1rk1/p4pp1/2pb3p/3p4/6Pr/1PNQ4/P1PB1PP1/4RRK1 b - - 0 1 > >solution is Bh2+ >Depth Time Nodes Score Princ. Continuation >----- ---- ----- ----- ------------------- >1 ... 0.00 226 -1.793 Rxg4 >2 ... 0.04 462 -1.997 Rxg4 Bf4 >3 ... 0.11 2970 199.991 Bh2+ Kh1 Bg3+ > >my move is Bh2+ >[right][3][0.108][26283][3.00] > > > >position 009 >3q1rk1/p4pp1/2pb3p/3p4/6Pr/1PNQ4/P1PB1PP1/4RRK1 b - - 0 1 > >solution is Bh2+ >Depth Time Nodes Score Princ. Continuation >----- ---- ----- ----- ------------------- >1 ... 0.01 411 199.991 Bh2+ Kh1 Bg3+ Kg1 Rh1+ Kxh1 Qh4+ Kg1 > Qh2# > >my move is Bh2+ >[right][1][0.014][15807][1.00] > > >--- >James I think that what really matters about extensions is nodes searched. In Tristram 4.0, it searched about 2x the number of nodes for ply 1, but got much better results; in your case I'd say the extensions are well worth it. You offer too few positions to judge whether or not you are overdoing the extensions (for instance, how does the pv look in a non-checking position?), so there isn't really anything I can say, other than I wish my program would solve that problem like that. :) Based on my experience with matching Tristram 3.36 against other programs, I think that it's major problem is missing simple tactical tricks that were just out of it's depth. More extensions (most of them were check situations) would really help. James
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.