Author: Ferdinand S. Mosca
Date: 05:57:46 03/20/01
Go up one level in this thread
On March 19, 2001 at 13:53:42, Dieter Buerssner wrote:
>On March 19, 2001 at 01:04:32, Ferdinand S. Mosca wrote:
>
>>Here is the log on P3 667
>[..]
>>I think it is similar.
>
>Thanks. Yes, looks almost identical.
>
>>On my new setting, Yace does not even consider Qe7, after Kg7, here it is:
>
>[...]
>> 66395034 3:50.9 -2.86 14t 1...Kg7 2. Qd7+ Kg6 3. h4 gxh4 4. Qe8+ Kg7 5.
>> Qe7+ Kh8 6. d7 Qc6+ 7. Kh2 Qc7+ 8. Kh3 Qc3+ 9.
>> Qe3 Qf6 10. Qe8+ Kh7 11. d8=R Qxf2 12. Qg8#
>> {-260}
>
>>The changes are:
>>icheck 0.6
>>irepcheck 1.0
>
>To readers not aware of Yace or its documantation. These are the extensions Yace
>does in case of check and "repeated check". Default values are 1.0 and 0.6.
>
>I am sorry, changing icheck (the fracional incncrement for search extensions
>when in check) won't work well right now. A value smaller 1.0 may cause an
>artifact seen above (and values bigger 1.0 get more or less truncated to 1.0 in
>most search situations). In the search routine, Yace assumes, that quiescence
>search will not be entered, when in check. But I think exactly this happens
>above, and you see a score of -2.8 for a line that ends with mate.
>So, I fear the use of icheck.
>
>Yes, this is a bug, and I will fix it.
>
>>My idea at first is to discourage Yace from making some checks, Yace is very
>>sensitive to checks, especially at endings, and I raise the value of irepcheck
>>on the idea that there is a serious bussiness when there is repetition of check.
>
>I agree with you, that Yace often does too many checking moves in endgames.
>
>However, decreasing check extensions (which Yace allready does in the background
>in certain endgame situations) my not yield in what you intend.
>
>Look at the following situation:
>
>1. almost any white move black advances passed pawn (PP)
Right but white has a heavy d-passed pawn, I have examined the position just
this afternoon. White's problem is to safeguard his king from perpetual check,
better to see the diagram.
[D]8/8/pq1PQ1kp/6p1/8/7P/5PK1/8 b - -
The line 1...Kg7 2 Qe7 Kg6 3 Qc7?? is a draw, 3...Qb5!= or 3...Qb3=
The move 2 Qe7 is the most normal square that can be seen easily, with the idea
of advancing the d6 pawn, but white has 2...Kg6 and if white will insist on
playing 3. d7 then black here has a forceful draw.
The move 3. Qc7 at this time is not good. When it is considered to be good? The
answer is very complicated but, it requires to move the h4 first before Qc7
becomes a good move.
By moving h4 (maybe hard for a computer, starting from the initial position) it
opens up the position of the black king. The idea is:
1. To provide the white queen some access squares when the time comes that both
queens are giving checks to each other, which proves to be advantageous to white
because there is a great chance for the queen to be swapped.
We will not fear black's passed a-pawn, it still very far, whereas White will
prepare at least 2-3 preparatory moves before advancing the d6 pawn.
2. At some point the h3 square vacated by the move h4 is actually a good square.
Here is my analysis on the behavior of Yace's and Crafty's evaluation of the
position. The following evaluations is from your first post.
>I tried this position with Yace. Something strange happened:
> 51305 0.702 -0.40 5. 1...Kg7 2.Qe7+ Kg6 3.Qe8+ Kf6 4.Qe7+ Kg6 5.Qe8+
> Kf6 6.Qe7+ Kg6 {0}
> 111931 1.638 -0.80 6-- 1...Kg7 2.Qe7+ Kg8 3.d7 Qc6+ 4.Kg3 Qc7+ 5.Kg4
> Qc6 6.d8=Q+ Qe8 {-921}
2...Kg8?? fail low, white can win this position easily.
> 137448 1.868 -7.64 6t 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.d8=Q
> Qxh3+ 6.Kg1 {-841}
> 150310 1.974 -7.64 6. 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.d8=Q
> Qxh3+ 6.Kg1 {-841}
>Here one could assume, that it really sees fast, that black is lost
> 186148 2.222 -7.75 7t 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.Qd6+ Kg7
> 6.Qe7+ Kg6 7.Qe4+H Kf7H 8.d8=QH Qxh3+H 9.Kg1H
> {-841}
> 191019 2.266 -7.75 7. 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.Qd6+ Kg7
> 6.Qe7+ Kg6 7.Qe4+ Kf7 8.d8=Q Qxh3+ 9.Kg1 {-841}
> 337668 3.199 -7.78 8t 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.Qd6+ Kf7
> 6.d8=Q Qxh3+ 7.Kg1 {-841}
> 505552 4.276 -7.78 8. 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 g4 5.Qd6+ Kf7
> 6.d8=Q Qxh3+ 7.Kg1 {-841}
>And it kept about the same for another 2 plies. But now:
> 1666011 12.245 -7.38 9++ 1...Kg7 2.Qe7+ Kg6 3.Qc7 Qb3 4.d7 Qd5+ 5.Kh2
> Qd4 6.d8=R Qxf2+ 7.Kh1 {-340}
Actually 3.Qc7?? is a forced draw. Fail high, and it is really a difficult
position to analyze but Yace here recognizes that it is a draw starting from
depth 6t till this current depth. 9++. Uri is right that there is a hidden
perpetual check from this position and this is after White's untimely 3. Qc7.
> 6934801 57.940 -0.40 9t 1...Kg7 2.Qd7+ Kg6 3.h4 {HT}
> 7368426 1:01.4 -0.40 9. 1...Kg7 2.Qd7+ Kg6 3.h4
From this point on White is guarateed a forced win. 1...Kg7 2.Qd7+ Kg6 3.h4!!
>Huch: we are back at almost drawing score.
This is just the begining negative score will continue to increase, good for
White.
> 8889435 1:13.5 -0.80 10-- 1...Kg7 2.Qd7+ Kg6 3.h4 gxh4 4.Qe8+ Kh7 5.Qe7+
> Kg6 6.d7 Qc6+ 7.Kh2 Qc7+ 8.Kh3 h5 9.d8=Q a5
> {-841}
> 12883052 1:46.1 -1.68 10t 1...Kg7 2.Qd7+ Kg6 3.h4 gxh4 4.Qe8+ Kf6 5.d7
> Qc6+ 6.Kh2 Qc7+ 7.Kh3 Qc3+ 8.Kxh4 Qc4+ 9.Kg3
> Qc7+ 10.f4 Qc3+ 11.Kg4 {HT} {0}
> 13257667 1:48.9 -1.68 10. 1...Kg7 2.Qd7+ Kg6 3.h4 gxh4 4.Qe8+ Kf6 5.d7
> Qc6+ 6.Kh2 Qc7+ 7.Kh3 Qc3+ 8.Kxh4 Qc4+ 9.Kg3
> Qc7+ 10.f4 Qc3+ 11.Kg4 {HT} {0}
> 17090908 2:15.6 -2.08 11-- 1...Kg7 2.Qd7+ Kg6 3.h4 gxh4 4.Qe8+ Kf6 5.d7
> Qc6+ 6.Kh2 Qc7+ 7.Kh3 Qc3+ 8.Kxh4 Qc4+ 9.Kg3
> Qc7+ 10.f4 h5 11.d8=Q+ Kf5 {-921}
> 26447712 3:26.8 -6.50 11t 1...Kg7 2.Qd7+ Kg6 3.h4 Kh5 4.Qe8+ Kxh4 5.Qe3
> Qxe3 6.fxe3 a5 7.d7 a4 8.d8=Q a3 9.e4 {-841}
> 26573747 3:27.8 -6.50 11. 1...Kg7 2.Qd7+ Kg6 3.h4 Kh5 4.Qe8+ Kxh4 5.Qe3
> Qxe3 6.fxe3 a5 7.d7 a4 8.d8=Q a3 9.e4 {-841}
> 33767907 4:20.6 -6.90 12-- 1...Kg7 2.Qd7+ Kg6 3.h4 Kh5 4.Qe8+ Kxh4 5.Qe3
> Qxe3 6.fxe3 a5 7.d7 Kh5 8.d8=Q Kg6
>And after rather long time again at loosing score. I have not investigated the
>reasons, but perhaps this is more complicated for a chess program, than it >looks
>at first sight.
Yace is correct.
Crafty's evaluation from my test on P3 667.
13-> 1:58 6.83 1. ... Kg7 2. Qe7+ Kg6 3. Qc7 Qb1 4.
d7 Qe4+ 5. Kh2 Qd4 6. Kg1 g4 7. d8=Q
Qxd8 8. Qxd8 gxh3
14 7:50 -- 1. ... Kg7
14 10:38 7.90 1. ... Kg7 2. Qd7+ Kg6 3. h4 g4 4.
Qxg4+ Kf6 5. Qf4+ Ke6 6. Qxh6+ Kd5
7. Qh5+ Kxd6 8. Qh6+ Kc5 9. Qxb6+ Kxb6
10. h5 a5 11. h6 a4
Yes it sees the correct sequence, 1. ... Kg7 2. Qd7+ Kg6 3. h4, but at long
time. Crafty has spend a lot of time detecting that Qc7 move is actually a draw
by perpetual check.
From my test on the new version
Yace's Default:
icheck 1.0
irepcheck 0.6
30816123 1:19.8 -1.39 11-- 1...Kg7 2. Qd7+ Kg6 3. h4 Qd4 4. Qe8+ Kg7 5.
Qe7+ Kg6 6. h5+ Kxh5 7. Qe8+ Kh4 8. Qe6 Kh5 9.
d7 g4 10. Qe8+ Kh4 11. d8=Q+ Qf6 {-841}
My setting:
icheck 0.6
irepcheck 1.0
8494473 21.774 -0.99 11t 1...Kg7 2. Qd7+ Kg6 3. h4 {HT}
My observation:
On the default, Yace does not consider 2. Qe7+ this is probably because it used
some extensions first (icheck 1.0) and optimized at the pv first that 2. Qd7 is
best which is absolutely correct. But it spends a lot of time examining 3. Qe8,
3. Qc7 and back again 3. Qe8 until at 3. h4! spending 1:19.8.
On my setting Yace sees immediately Qe7, ignoring first the move Qd7 (icheck
0.6) which is very natural, not blockading the passed d-pawn, just like what
crafty does. But it easily sees back the move 2. Qd7 followed up by h4, or maybe
detected the move Qc7 to be a draw earlier. 21.774 only.
From my point of view as a player, I usually make some logical variation on
mind, (but not spending a real time analyzing the first best move) then modify
the sequence/moves.
Probably the default tries to determine the best first move 2. Qd7 and
completely ignore Qe7, while my setting analyzes the move Qe7 like Crafty and
found out later that Qd7 is the best by visiting a lot of variations leading to
the early discovery of h4.
Thanks for some info.
Regards,
Dinan
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.