Author: Robert Hyatt
Date: 08:55:07 11/22/05
Go up one level in this thread
On November 22, 2005 at 08:52:09, Kolss wrote:
>On November 22, 2005 at 07:19:24, Bernhard Bauer wrote:
>
>>It takes quite a lot of time for TogaII 1.1 to solve this simple position.
>>Why? And what can be done to become faster?
>>
>>[D]8/7p/8/p4p2/5K2/Bpk3P1/4P2P/8 w
>
>
>Hi,
>
>That is a good question. Originally, I thought that tablesbases (TB) and / or
>bitbases (BB) would help in this position, as some of the critical positions in
>the tree are 5-piece endings. However, when I let Ikarus run without TB or BB,
>or with TB (5-piece) but no BB, or with TB and BB (only 4-piece), or with TB and
>BB (4-piece plus PP-P), I essentially always get virtually the same results. In
>particular, the solution (g4!) is always found in ply 11, and the big fail-high
>(to > +3) always occurs in ply 12.
>
>Therefore, I guess that it is most likely mainly due to the evaluation of pawn
>endgames and the queen endgame in the other critical variation, although the TB
>certainly don't hurt much as well.
>
>The main critical line is:
>
>1. g4 fxg4 2. e4 b2 3. Bxb2+ Kxb2 4. e5 a4 5. e6 a3 6. e7 a2 7. e8=Q a1=Q 8.
>Qe5+ (or Qh8+) Kb1 {side comment: the rest could be covered by quiescence
>search} 9. Qxa1+ Kxa1 10. Kxg4.
>
>[D]8/7p/8/8/6K1/8/7P/k7 b - - 0 10
>
>This is a 4-piece pawn ending. If a program evaluates this highly in favor of
>white (because of the distant black king), it will find the solution easily. If
>not, it is much more tricky, as the pawn on h7 will only be captured another 6
>plies later (10. ... Kb2 11. Kh5 Kc3 12. Kh6 Kd4 13. Kxh7). Now that is 25 (24)
>plies down the tree before any program should see a clear advantage for white
>(side comment: which however can be dangerous, as black is lacking only one
>tempo to achieve the draw).
>
This is something I had to add a long while back, otherwise an outside passed
pawn evaluation fails because you have to use the pawn as a decoy to get the
opponent's king too far away, as here.
Crafty's static eval here:
material evaluation................. 0.00
development......................... 0.00
pawn evaluation..................... 0.00
passed pawn evaluation.............. 0.00
passed pawn race evaluation......... 0.00
interactive piece evaluation........ 2.00
total evaluation.................... 2.00
>One of the other important lines is:
>
>1. Kxf5 b2 2. Bxb2+ Kxb2 3. e4 a4 4. e5 a3 5. e6 a2 6. e7 a1=Q 7. e8=Q.
>
>[D]4Q3/7p/8/5K2/8/6P1/1k5P/q7 b - - 0 7
>
>Note that here, black queened first and now can give some annoying checks. If I
>let Ikarus run in this position, the score stays stable for quite a while
>(several minutes) at around +1.3 to +1.4.
>
>Depending on how the respective programs evaluate the positions in diagrams 1
>and 2, it may be easy or very difficult to find the correct move.
>
>Of course, this is not at all an encompassing analysis, but I hope that helps a
>bit to understand what might be going on.
>
>Best regards - Munjong.
much of endgame skill is more about knowledge than search. The things are often
too deep, particularly when you search deeply to reach the position in question,
and you really want to evaluate it as "won" so that you will try to reach it,
rather than just knowing how to win it if you ever do get there because you have
search tricks to selectively search deep enough (from the root) to spot it in
the required time...
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.