Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / a reverse?
- - By lech (*) Date 2010-02-10 18:40
Is there an engine which can find the simple winning move 1...Qxh7 ?
8/5pKN/5Pp1/5kP1/1B6/8/b7/7q b - - 0 1

Can anyone test it?
Parent - - By BrandonSi (***) Date 2010-02-10 20:09
My brain is not working very well today, so apologies for the silly question, but from a quick glance at Qxh7, are you sure it's a 100% winning position? I see a few holes, but I only looked at it for a minute.
Parent - By Carl Bicknell (*****) Date 2010-02-10 20:12
seems like a win to me. White will have to give up his bishop to stop the g pawn and then it's easy.
Parent - By Kkuo (**) Date 2010-02-18 18:21 Edited 2010-02-18 18:25
Its just typical fortress positions that computers can't recognize.  Most are not practical positions.
- - By lech (*) Date 2010-02-10 20:17
And what can top-egines play?
Parent - - By BrandonSi (***) Date 2010-02-10 20:24
Rybka like this..
  -+  (-6.86)   Depth: 18   00:00:23  1089kN

Spark prefers..
1...Qb7 2.Bc5 Qc7 3.Be3 Qh2 4.Bc1 Qh4 5.Kg8 Bc4 6.Be3 Qg3 7.Bc5 Qe5 8.Bg1 Bd5 9.Ba7 Qh2 10.Be3 Qg3 11.Bd4 Qb8+ 12.Kg7 Qc8 13.Be3 Qa8 14.Bd4 Qe8 15.Kh6
  -+  (-9.35)   Depth: 26/34   00:00:13  19523kN

Another engine saw this..
1...Qd5 2.Be1 Qd4 3.Bg3 Bd5 4.Bh2 Qe3 5.Bc7 Kg4 6.Bd6 Qd4 7.Bh2 Qa7 8.Bd6 Qe3 9.Bb4 Qd4 10.Be1 Kf4 11.Ba5 Qe3 12.Bb4 Kf5 13.Bd6 Qc3 14.Bb8 Kg4 15.Ba7 Kf4 16.Bf2
  -+  (-7.75)   Depth: 29   00:00:16  20010kN
Parent - - By lech (*) Date 2010-02-10 20:50
What ELO should such engines have?
Parent - By BrandonSi (***) Date 2010-02-10 20:54
A higher ELO than me :)

I think Rybka 3 is somewhere around 3200-3250.. Spark I believe is 3000. The other engine is supposed to be ~3300..
Parent - - By Uly (Gold) Date 2010-02-11 02:50
Do you draw against these engines (allowing takebacks) after their suggested moves?
Parent - - By BrandonSi (***) Date 2010-02-11 03:34
Not sure I understand the question, Vytron?

I often find it speeds up the depth search to make the suggested move, then retract that move. From my understanding, doing so stores the partial results of that first actual (recommended) move in the hash table. For me it's a quick sanity check.. Telling the engine, "OK, now that you've made the move and partially increased the depth of that particular move/line, lets back up and re-evaluate the alternatives."

I assume you ask, as my kn/s are obviously on the high side?

If you're asking if I actually play chess against these engines, I do not. In reality I'm not a very good chess player. :)
Parent - By Uly (Gold) Date 2010-02-11 06:13
Sorry for the misunderstanding, I was asking because the thread seems to be about the move Qxh7 and how the engines want to play something else, however, I wouldn't see a problem if the moves suggested by the engines win too, and I don't even see if Qxh7 is a faster win.

I guess the question was directed at lech, but you posted the engines' moves.
- - By lech (*) Date 2010-02-11 13:06
Moves suggested by engines don’t win, of course.
Since the simple move 1...Qxh7 would make a beginning chess player, all engines, which are not able to find this move, doesn’t deserve a serious treatment and high ELO.
The creators of engines should immediately improve their software. In other case, the payment for commercial products should be returned. 
For people, it is not so important that an engine is winning computer matches continually. The engine should be reliable one.
Parent - - By jackwelb (*) Date 2010-02-11 15:46
The moves suggested by computers will eventually win after the 50-move rule forces a black queen sacrifice on g5 or h7.  While this may be the best use of time in game-mode, chess engines should do better in analyze-mode.  (BTW, what does the UCI option “analyze-mode” do anyway?)

Now if someone could come up with a chess engine that actually analyzed a position, rather than minmax, that would be something.  Such as a series of statements:
“Black holds a large material advantage.”
“The pawns are locked. Further pawn progress would require a piece capture.”
“The white king cannot be forced to move without a black queen sacrifice.”

I can only imagine how difficult this would be to program since we have a 1000 chess engines and not 1 which can state the strengths and weaknesses in a position.

PS - Is there a way to set-up a position where you are on move 45 of the 50-move rule?  Chess engines would then suggest much different moves in the above position.
Parent - - By lech (*) Date 2010-02-11 16:11
It is possible to produce such positions (other statements).
rk6/p1p5/Bp2p3/1P6/q2B1P2/4K3/8/8 b - - 0 1

wins Qxa6 (Qxb5 or Qxd4)
Maybe, it is difficult to improve algorithms, but it is impossible to abandon this question.
Parent - - By lech (*) Date 2010-02-11 16:31
In the first position, white Bishop can take a better place  b4->f8->h6 and 50 moves dont help to force black Queen to capture.
Parent - - By lech (*) Date 2010-02-11 17:07 Edited 2010-02-11 17:31
It seems to me that if it is possible to produce such positions, it should also be possible to handle this question. We should force programmers to do it. In other case each analyze or play can contain such a critical error.
Parent - - By Uly (Gold) Date 2010-02-12 02:24
Do these positions happen in actual games?
Parent - - By lech (*) Date 2010-02-12 21:03 Edited 2010-02-12 21:06
I found two next positions and now it should be easy to call and, perhaps, solve the problem.

7k/6rB/6P1/8/2p5/2B2q2/3P4/4K3 b - - 0 1

wins Qxc3
6R1/5Bpr/2q3Pk/8/8/8/8/3K4 b - - 0 1

wins Kg5, now or later

I called this error “crazy queen”. In all the positions, there is a wrong high score, and the queen makes a sequence of moves without any benefit.
It is hard to say something more about the problem. “Crazy queen” needs some specific conditions to arise, though.
Now, I better know the problem and I think that it is a very rare event, but, of course, may happen in play.
Parent - - By lech (*) Date 2010-02-13 22:14
I joined all my work into the article: "Crazy queen" and other problems with chess engines.
Parent - By lech (*) Date 2010-02-18 05:28
Unfortunately, it is possible to arise a crazy rook too.
7k/6rp/6pN/6P1/8/2B5/6r1/K7 b - - 0 1

1…Rxg5 wins
Up Topic The Rybka Lounge / Computer Chess / a reverse?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill