Author: Martin Giepmans
Date: 12:12:10 01/11/02
Go up one level in this thread
On January 11, 2002 at 13:09:24, David Rasmussen wrote: >On January 11, 2002 at 11:44:39, David Rasmussen wrote: > >>[D]2Q2n2/2R4p/1p1qpp1k/8/3P3P/3B2P1/5PK1/r7 w - - 0 1 >> >>In the above position, my search tree explodes because of extensions. It takes > >>3 minutes and 42 million nodes to finish an 8 ply search. How does your program >>do? >> >>Any good ideas on how to limit extensions in such a position. >> >>P.S. Solving the position is no problem. It is a simple mate in 5 plies, and is >>found very early on at depth 1 or 2 or so. Still, there must be something >>unsound about my extensions, or at least room for improvement, when this >>position makes the tree explode. >> >>/David > >The same thing happens with WAC 161: > >[D]3r3k/3r1P1p/pp1Nn3/2pp4/7Q/6R1/Pq4PP/5RK1 w - - 0 1 > >time nodes depth score pv >00:00:00.17 33kn 05/16/16 --------------------------------------------------- >00:00:00.19 36kn 06/12/12 +MAT7 1. Qxd8+ Rxd8 2. f8=Q+ Rxf8 > 3. Rxf8+ Nxf8 4. Nf7# >00:00:00.56 122kn 06/17/17 --------------------------------------------------- >00:00:00.59 132kn 07/13/13 +MAT7 1. Qxd8+ Rxd8 2. f8=Q+ Rxf8 > 3. Rxf8+ Nxf8 4. Nf7# >00:00:07.75 1887kn 07/46/46 --------------------------------------------------- >00:00:07.80 1906kn 08/14/14 +MAT7 1. Qxd8+ Rxd8 2. f8=Q+ Rxf8 > 3. Rxf8+ Nxf8 4. Nf7# >00:03:25.14 51Mn 08/54/54 --------------------------------------------------- >00:03:25.19 51Mn 09/15/15 +MAT7 1. Qxd8+ Rxd8 2. f8=Q+ Rxf8 > 3. Rxf8+ Nxf8 4. Nf7# >00:47:45.77 736Mn 09/62/62 2/52 1. Ne8 > >This is while searching. The search is not finished. Hence the last line showing >only the move being searched right now. The numbers in the "depth" row (x/y/z) >are nominal search depth, max depth at which q-search begins (this can be higher >due to extensions), and max depth at which q-search ends. From the numbers >above, you can see that for ply 8 and for ply 9 up until now, we extend in the >longest line to 54 and 62 plies respectively! That is too much. My extensions >right now are: >Check extension: 1 ply >Pawn push extension to rank 7 (2): 0.75 ply >Recapture extension: 0.75 ply >One Reply extension: 0.75 ply > >The sum of extensions at a ply are limited to at most one ply. Although one >reply extension is added _after_ this limiting, as one_reply wouldn't make any >difference at all. I mean, it is only applied in check positions where there is >only one move, and in those positions I already add the check extension. So 1 + >0.75 = 1.75, and limiting this to 1 ply would yield 1 ply. How do you guys >tackle this? It's probably one of those small bugs that are so easy to overlook. There is a pattern: just before ----- (end of iteration?) the count is way too high. After that (first move of next iteration) it seems ok again. So first rootmove ok, second and next move probably not. Something wrong with initialising the extension-counter when returning to the root?? Martin
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.