Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Problems finding mates?
- - By Arifur (**) [gb] Date 2008-09-26 11:34
In this position Rybka 3 have problems seeing the Rxc3 (and instead goes for Nd7?) and mate in 7 according to Zappa (and other engines). I let Rybka think for 8 minutes, but could not find Rxc3 (not shown below). I have noticed also in some other positions Rybka have problems finding mates, other programs finds mates much quicker.

New game -
1Q6/2r1kppp/4p3/3pn3/1P3q2/P1P4P/6P1/7K b - - 0 1


Analysis by Rybka 3:

31...Nd7
  -+  (-17.28)   Depth: 6   00:00:00  20kN
31...Nd7
  -+  (-18.28)   Depth: 7   00:00:01  162kN
31...Nd7
  -+  (-18.49)   Depth: 8   00:00:01  198kN
31...Nd7
  -+  (-19.44)   Depth: 9   00:00:11  2693kN
31...Nd7
  -+  (-19.87)   Depth: 10   00:00:19  4775kN
31...Nd7
  -+  (-#18)   Depth: 11   00:01:40  25079kN
31...Nd7
  -+  (-#18)   Depth: 12   00:01:51  27734kN
31...Nd7
  -+  (-#15)   Depth: 13   00:01:57  29088kN

(, 26.09.2008)

Analysis by Zappa Mexico II:

31...Rxc3 32.Qa7+ Kf6 33.Qg1 Rc1 34.Qxc1 Qxc1+ 35.Kh2 Qxa3
  -+  (-15.64)   Depth: 8/22   00:00:01  41kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7
  -+  (-17.21)   Depth: 8/22   00:00:01  53kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7
  -+  (-17.21)   Depth: 8/22   00:00:02  66kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 9/31   00:00:03  155kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 9/31   00:00:04  198kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 10/31   00:00:04  225kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 10/31   00:00:04  355kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 11/31   00:00:04  473kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 11/31   00:00:05  783kN
31...Qf1+ 32.Kh2 Nf3+ 33.gxf3 Qf2+ 34.Kh1 Rxc3 35.Qa7+ Qxa7 36.Kg2 Qe3 37.Kg3 Qxf3+ 38.Kh2 Rc2+ 39.Kg1 Qg2#
  -+  (-#9)   Depth: 12/31   00:00:05  1074kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 12/31   00:00:05  1222kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 12/31   00:00:06  1409kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 13/31   00:00:06  1627kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 13/31   00:00:07  2416kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 14/31   00:00:07  2750kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 14/31   00:00:08  4040kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 15/35   00:00:09  4690kN
31...Rxc3 32.Qa7+ Kf6 33.Qc5 Rxc5 34.bxc5 Nc4 35.Kg1 Ne3 36.Kh1 Qf1+ 37.Kh2 Qxg2#
  -+  (-#7)   Depth: 15/35   00:00:12  8086kN

(, 26.09.2008)
Parent - - By Arsha (**) Date 2008-09-26 12:24 Edited 2008-09-26 12:30
Rybka 2.3.2a was finding mates better.

Using infinite analysis, she could find Rxc3 instead of Nd7!
However it is very SLOW.
Parent - - By Carl Bicknell (*****) [gb] Date 2008-09-26 12:40
I wonder why because R3 is much stronger tactically than R2.3.2a
Parent - By Vempele (Silver) [fi] Date 2008-09-26 14:26
Rybka 3 doesn't hash mate scores.
Parent - - By Gr00vy (**) [us] Date 2008-09-26 14:51
This should be expected, and probably desirable behavior from Rybka 3.  In it's selective search it is going to be narrow and deep.  It found a Mate, much deeper than the one that is best, but it has absolutely no reason to go and make the tree wider, a mates a mate.

Now granted, if it was given extra time, it could search for quicker mates.  This would be a much more interesting case, if you could create a construct where there was *not* an alternate and longer mate available. 

But, that said, in a non-exhaustive search, you can always create a case where the selectivity fails.  Always, by definition.  This doesn't make the engine "weak", not by any means.  It's selectivity is key to it's strength.  Just that no matter how big a deal the practical upside of a selective search, there will be by definition, some unusual winning moves that it will miss.
Parent - By Kedar (*) [cz] Date 2008-11-16 09:42
Your theory is very interesting and probably it is very close to real Rybka 3 setup by the programmers.
But still I do not see the point. I would expect that a new version of a program should to take over all the good algorithms from it's predecessors. Of course - new versions have new ideas, but in this case I am not able to understand why Vas and his team went this way. Mate is the main goal in chess and I think it should be achieved the fastest way. For me it looks like this - living in New York, why should I buy my newspapers around the corner when I can buy them in Los Angeles.
Parent - By sovi (**) [cz] Date 2008-10-02 05:22 Edited 2008-10-02 05:28
Which engine is seeing mate ?
8/1Q3pkp/4p1p1/1p2P3/8/R6P/1r3qP1/7K b - - 0 47
Parent - By Zherkovv (***) Date 2008-11-15 10:14
New game, 30"/1

Mate in 2743 ???  WOW

1Q6/2r1kppp/4p3/3pn3/1P3q2/P1P4P/6P1/7K b - - 0 1


Analysis by Movei00_8_438:

1...Rxc3 2.Qb7+ Kf6 3.Qa7 Rc1+ 4.Qg1 Rxg1+ 5.Kxg1
  -+  (-16.02)   Depth: 6   00:00:00  16kN
1...Rxc3 2.Qb7+ Kf6 3.Qa7 Rc1+ 4.Qg1 Rxg1+ 5.Kxg1
  -+  (-16.02)   Depth: 6   00:00:00  20kN
1...Rxc3
  -+  (-16.32)   Depth: 7   00:00:00  23kN
1...Rxc3
  -+  (-17.02)   Depth: 7   00:00:00  24kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qxa3
  -+  (-17.80)   Depth: 7   00:00:00  30kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qxa3
  -+  (-17.80)   Depth: 7   00:00:00  36kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qb2 6.Kg1 Qxa3
  -+  (-17.56)   Depth: 8   00:00:00  44kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qb2 6.Kg1 Qxa3
  -+  (-17.56)   Depth: 8   00:00:00  60kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qxa3 6.b5 Kf5 7.g4+ Ke4
  -+  (-17.61)   Depth: 9   00:00:00  74kN
1...Rxc3 2.Qa7+ Kf6 3.Qg1 Rc1 4.Qxc1 Qxc1+ 5.Kh2 Qxa3 6.b5 Kf5 7.g4+ Ke4
  -+  (-17.61)   Depth: 9   00:00:00  119kN
1...Rxc3
  -+  (-17.91)   Depth: 10   00:00:00  134kN
1...Rxc3
  -+  (-18.61)   Depth: 10   00:00:00  138kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 Rxc5 4.bxc5 d4 5.Kg1 d3 6.g3 Qe3+ 7.Kf1 Qe2+ 8.Kg1
  -+  (-22.36)   Depth: 10   00:00:00  251kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 Rxc5 4.bxc5 d4 5.Kg1 d3 6.g3 Qe3+ 7.Kf1 Qe2+ 8.Kg1
  -+  (-22.36)   Depth: 10   00:00:01  284kN
1...Rxc3
  -+  (-22.66)   Depth: 11   00:00:01  393kN
1...Rxc3
  -+  (-23.36)   Depth: 11   00:00:01  418kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 d4 4.a4 Rxc5 5.bxc5 Qc1+ 6.Kh2 d3 7.a5 d2 8.a6 d1Q
  -+  (-25.07)   Depth: 11   00:00:02  615kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 d4 4.a4 Rxc5 5.bxc5 Qc1+ 6.Kh2 d3 7.a5 d2 8.a6 d1Q
  -+  (-25.07)   Depth: 11   00:00:02  730kN
1...Rxc3
  -+  (-24.77)   Depth: 12   00:00:02  831kN
1...Rxc3
  -+  (-24.07)   Depth: 12   00:00:03  1142kN
1...Rxc3
  -+  (-25.37)   Depth: 12   00:00:03  1207kN
1...Rxc3
  -+  (-25.37)   Depth: 12   00:00:04  1575kN
1...Rxc3
  -+  (-25.67)   Depth: 13   00:00:05  1693kN
1...Rxc3
  -+  (-26.37)   Depth: 13   00:00:05  1827kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 Rxc5 4.bxc5 d4 5.Kg1 d3 6.g4 Qe3+ 7.Kh1 d2 8.g5+ Kxg5 9.h4+ Kxh4 10.Kh2 Ng4+ 11.Kh1 Qf3+ 12.Kg1 d1Q#
  +-  (#2743)   Depth: 13   00:00:07  2555kN
1...Rxc3 2.Qa7+ Kf6 3.Qc5 Rxc5 4.bxc5 d4 5.Kg1 d3 6.g4 Qe3+ 7.Kh1 d2 8.g5+ Kxg5 9.h4+ Kxh4 10.Kh2 Ng4+ 11.Kh1 Qf3+ 12.Kg1 d1Q#
  +-  (#2743)   Depth: 13   00:00:08  2895kN

(,  15.11.2008)
Up Topic Rybka Support & Discussion / Rybka Discussion / Problems finding mates?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill