Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Deep rybka cannot solve a position I solved in 10 min. WTF??
- - By AlKa Date 2009-08-13 01:02
Hi. I'm totally new to computer chess and only got deep rybka a week ago. It's actually pretty interesting to play through super GMs' games with the engine. Anyhow, here's the position DR3 couldn't solve: White: Kd5, Qb8, Rf1, Nf3, pawn b2. Black Kh1, Qh3, Bg1, pawns h5, h2, g2, c5. White to move. I ran it on 2.53 Ghz, 4gb ddr3 ram (2 gb hash for the engine), Vista 64-bit for over an hour. I'm interested to see how DR would do with quad core i7 desktop trying to solve this position (I'm thinking about getting one). I really would like to know if DR would play much stronger on quad core i7 than on my laptop. If you try this position on your PC, please post system specs. with your reply. Thanks.
Parent - - By Uly (Gold) Date 2009-08-13 01:11

> here's the position DR3 couldn't solve:


Details please, what is the goal that the best move achieves on the position and does Rybka's move not achieve the same goal?
Parent - - By AlKa Date 2009-08-13 01:52
Sorry. It' white to move and win ( not just 1 move but a variation that leads to a clearly winning position for white). DR3  sees the right move but gives the wrong evaluation because it does not see the win.
Parent - - By Uri Blass (*****) Date 2009-08-13 03:51
Rybka3 on a slow hardware give the right evaluation but cannot show more than 5.12 pawns advantage(I used only 256 mbytes hash
and a single processor machine with no tablebases)

1...Qxh2 is the only legal move but rybka cannot show it in the pv(there is no justification not to continue main line when there is a single legal move and there is no justification not to show a move after Qxh2 that rybka must have in her hash tables so I think that there is a bug in the program that prints the pv because it does not show the pv
New game
1Q6/8/8/2pK3p/8/5N1q/1P4pp/5Rbk w - - 0 1


Analysis by Rybka 3 1-cpu 32-bit :

1.Qb8xh2+
  -+  (-3.24)   Depth: 2   00:00:00
1.Qb8xh2+
  -+  (-3.80)   Depth: 3   00:00:00
1.Qb8xh2+
  -+  (-3.60)   Depth: 4   00:00:00
1.Qb8xh2+
  -+  (-3.40)   Depth: 4   00:00:00
1.Qb8xh2+
  -+  (-3.00)   Depth: 4   00:00:00
1.Qb8xh2+
  -+  (-3.31)   Depth: 4   00:00:00
1.Qb8xh2+ Qh3xh2
  -+  (-3.11)   Depth: 5   00:00:00
1.Qb8xh2+ Qh3xh2 2.Rf1-a1
  -+  (-3.98)   Depth: 5   00:00:00
1.Qb8xh2+
  -+  (-3.78)   Depth: 6   00:00:00  3kN
1.Qb8xh2+
  -+  (-3.58)   Depth: 6   00:00:00  3kN
1.Qb8xh2+
  -+  (-3.18)   Depth: 6   00:00:00  3kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4
  -+  (-3.91)   Depth: 6   00:00:00  3kN
1.Qb8xh2+
  -+  (-3.71)   Depth: 7   00:00:00  4kN
1.Qb8xh2+
  -+  (-3.51)   Depth: 7   00:00:00  4kN
1.Qb8xh2+
  -+  (-3.11)   Depth: 7   00:00:00  4kN
1.Qb8xh2+
  -+  (-2.31)   Depth: 7   00:00:00  4kN
1.Qb8xh2+
  µ  (-0.71)   Depth: 7   00:00:01  4kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 h4-h3 4.Ra1-e1
  =  (0.00)   Depth: 7   00:00:01  5kN
1.Qb8xh2+
  =  (0.20)   Depth: 8   00:00:01  6kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 h4-h3 4.Ra1-e1 c5-c4 5.Ke4-d5 Qh2-e5+ 6.Kd5xe5 h3-h2 7.Ke5-d5 c4-c3 8.Nf3xh2
  =  (0.25)   Depth: 8   00:00:01  6kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Ke4-d5 Qh2-d6+ 6.Kd5xd6 h3-h2 7.Kd6-d5 c4-c3 8.Nf3xh2 c3xb2 9.Re1-b1 Kh1xh2 10.Rb1xb2
  =  (0.04)   Depth: 9   00:00:01  11kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Ke4-d5 Qh2-d6+ 6.Kd5xd6 h3-h2 7.Kd6-d5 c4-c3 8.Nf3xh2 c3xb2 9.Re1-b1 Kh1xh2 10.Rb1xb2 Kh2-g3
  =  (0.04)   Depth: 10   00:00:01  13kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Ke4-d5 Qh2-d6+ 6.Kd5xd6 h3-h2 7.Kd6-d5 c4-c3 8.Nf3xh2 c3xb2 9.Re1-b1 Kh1xh2 10.Rb1xb2 Kh2-g3
  =  (0.04)   Depth: 11   00:00:01  18kN
1.Qb8xh2+
  =  (0.24)   Depth: 12   00:00:01  28kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Ke4-d5 c4-c3 6.b2xc3 Qh2-d6+ 7.Kd5xd6 h3-h2 8.Re1-e2 Bg1-b6 9.Kd6-c6 Bb6-e3 10.Kc6-d5
  =  (0.11)   Depth: 12   00:00:01  42kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Ke4-d5 c4-c3 6.b2xc3 Qh2-d6+ 7.Kd5xd6 h3-h2 8.Re1-e2 Bg1-b6 9.Kd6-c6 Bb6-e3 10.Kc6-d5
  =  (0.05)   Depth: 13   00:00:02  54kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Re1-a1 c4-c3 6.b2xc3 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 Bb6-c7+ 10.Ke5-e4 Bc7-d6 11.Rh8-g8
  =  (0.07)   Depth: 14   00:00:02  91kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Re1-a1 c4-c3 6.b2xc3 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 Bb6-c7+ 10.Ke5-e4 Bc7-d6 11.Rh8-g8
  =  (0.07)   Depth: 15   00:00:03  114kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1
  =  (0.07)   Depth: 16   00:00:04  151kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1
  =  (0.01)   Depth: 17   00:00:06  294kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1
  =  (0.01)   Depth: 18   00:00:07  389kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Re1-a1 c4-c3 6.b2xc3 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 g2-g1Q 10.Nf3xg1 Kh1xg1 11.Ke5-e4 h2-h1Q+ 12.Rh8xh1+ Kg1xh1 13.c3-c4 Kh1-g2 14.Ke4-e5 Kg2-g3
  =  (0.01)   Depth: 19   00:00:09  578kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Re1-a1 c4-c3 6.b2xc3 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 g2-g1Q 10.Nf3xg1 Kh1xg1 11.Ke5-d5 h2-h1Q+ 12.Rh8xh1+ Kg1xh1 13.Kd5-c4 Bb6-g1 14.Kc4-b4 Kh1-h2
  =  (0.01)   Depth: 20   00:00:14  977kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 h4-h3 5.Re1-a1 c4-c3 6.b2xc3 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 g2-g1Q 10.Nf3xg1 Kh1xg1 11.Ke5-f4 h2-h1Q 12.Rh8xh1+ Kg1xh1 13.Kf4-g3 Bb6-c5 14.c3-c4 Kh1-g1 15.Kg3-f3 Kg1-h2
  =  (0.01)   Depth: 21   00:00:22  1488kN
1.Qb8xh2+ Qh3xh2 2.Rf1-a1 h5-h4 3.Kd5-e4 c5-c4 4.Ra1-e1 c4-c3 5.b2xc3 h4-h3 6.Re1-a1 Qh2-e5+ 7.Ke4xe5 h3-h2 8.Ra1-a8 Bg1-b6 9.Ra8-h8 g2-g1Q 10.Nf3xg1 Kh1xg1 11.Ke5-f4 h2-h1Q 12.Rh8xh1+ Kg1xh1 13.Kf4-g3 Bb6-c5
  =  (0.01)   Depth: 22   00:00:47  3165kN
1.Qb8xh2+
  =  (0.21)   Depth: 23   00:02:46  10389kN
1.Qb8xh2+
  ²  (0.41)   Depth: 23   00:02:51  10672kN
1.Qb8xh2+
  ±  (0.81)   Depth: 23   00:02:57  11013kN
1.Qb8xh2+
  +-  (1.61)   Depth: 23   00:03:05  11459kN
1.Qb8xh2+
  +-  (3.21)   Depth: 23   00:03:12  11883kN
1.Qb8xh2+
  +-  (5.12)   Depth: 23   00:22:18  71924kN

(so k,  13.08.2009)

possible line based on forcing rybka to play for both sides is:

[Event "?"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "New game"]
[Black "?"]
[Result "*"]
[SetUp "1"]
[FEN "1Q6/8/8/2pK3p/8/5N1q/1P4pp/5Rbk w - - 0 1"]
[PlyCount "19"]

1. Qxh2+ Qxh2 2. Rb1 c4 3. Kc6 h4 4. Kb7 h3 5. Ka8 c3 6. bxc3 Qb8+ 7. Rxb8 Bc5
8. Rh8 g1=N 9. Nxg1 Kxg1 10. Rxh3 *
Parent - - By AlKa Date 2009-08-13 05:50
You got pretty much the right line, Uri Blass. But how is it possible that your rybka gets the right evaluation on a single processor in 22 minutes at depth 23 when my duo core reaches depth 14 in over an hour??? ( and the highest kn/s I've seen on my system is approx. 300, deffinately not 71924!!!) BTW, Permanent Brain got to depth 28 in under 5 minutes (didn't solve it though). Are there some settings that I need to adjust?? What's going on??
Parent - - By nine castles (**) Date 2009-08-13 06:12
You probably have some process(es) in the background that are chewing up all your CPU time.
Parent - By AlKa Date 2009-08-13 06:22
No, I have nothing running in the background; CPU usage before I turn rybka on is about 0-3%.
Parent - By yanquis1972 (****) Date 2009-08-13 06:46
those are total nodes, not kn/s -- on a dual kn/s of 300 sounds very good to me. depth 14 after an hour is horrible. either you are mistaken here or you had a LOT of processes running. on a dual rybka should reach d14 very quickly.  in this position probably instantly.

to answer the original question, you WILL notice a big difference if you upgrade to an i7.  this comes down to your personal circumstances, needs, & patience (ie in 6 months or so we'll have the i9 & nehalem EX etc), but you can trust you will notice a huge difference. when i got my quad i thought it would be all i ever need, & that id never really have any use for all the cores for anything other than as a total luxury, but now i never analyze w/ anythign other than a rybka personality w/ all cores available & often wish i could get into depth 18-21 range in the opening/early middle positions much much faster.
Parent - By Uly (Gold) Date 2009-08-13 22:07

> DR3  sees the right mov


Rybka is designed to find the best move, that doesn't guarantee that the PV contains the strongest variation, i.e. the expected reply could fall immediately as soon as you force the move and you get a more accurate line.

In this sense, the position can be considered solved if Rybka finds the best move (regardless if other people on this thread found that Rybka did find the best line) even if "Rybka solves the position for the wrong reasons."
Parent - - By Gaмßito (****) Date 2009-08-13 02:17
Very interesting. It seems to be a zugzwang position.

Here is:

1Q6/8/8/2pK3p/8/5N1q/1P4pp/5Rbk w - - 0 0


> I really would like to know if DR would play much stronger on quad core i7 than on my laptop.


Of course it will play stronger. There are no Laptops faster than an i7! :-)

Regards,
Gaмßito.
Parent - - By AlKa Date 2009-08-13 02:35
I know, but how much better is DR on i7 than intel duo? Will is work 2 or 3 times as fast?
Parent - By Gaмßito (****) Date 2009-08-13 03:00
Between 3 and 4 times faster.

You can check that precisely here: http://www.sedatchess.com/rybkabenchmarks.html

Regards,
Gaмßito.
Parent - By mocha1961 (***) Date 2009-08-13 13:13
there are laptops on i7 now. FYI
Parent - By h1a8 (***) Date 2009-08-14 05:48
My laptop is. Just let Jessica Biel sit on it.
Parent - By Permanent Brain (*****) Date 2009-08-13 04:31
What exactly do you mean by solve, the correct first move (only), or a winning evaluation too, or even a perfect principle variation?

Anyway this is a quite unusual position, if I see it correctly with several zugzwangs involved during the solution (often a typical problem for engines). But it looks like Rybka 3 goes for Qxh2+ immediatly, and so does R2.2n2, but up to depth 28 it does not see a win.

Analysis by Rybka 2.2n2 mp 32-bit an-ftb, D945 (~dual P4)/3.4 GHz, 256 MB hash:

1.Qxh2+ Qxh2 2.Rb1 h4 3.Ra1 c4 4.Rc1 h3 5.Ke4 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Rc2 Bb6 9.Rc1+ Bg1 10.Rc2 Bb6 11.Rc1+ Bg1 12.Rc2 Bb6 13.Rc1+ Bg1 14.Rc2 Bb6 15.Rc1+ Bg1 16.Rc2 Bb6
  -+  (-1.66)   Depth: 2   00:00:00
  =  (0.00)   Depth: 28   00:04:45  57900kN
Parent - By Dadi Jonsson (Silver) Date 2009-08-13 09:14
Here are the results when I run this on my octal in Aquarium. The score is +3.2 after a minute and goes up to +6.4 two minutes later:

[+6.40]  d=22  1.Qxh2+ (0:02:39) 26837kN
[+3.20]  d=22  1.Qxh2+ (0:00:59) 3159kN
[+1.60]  d=22  1.Qxh2+ (0:00:57) 3151kN
[+0.80]  d=22  1.Qxh2+ (0:00:56) 3134kN
[+0.40]  d=22  1.Qxh2+ (0:00:52) 2999kN
[+0.20]  d=22  1.Qxh2+ (0:00:50) 2984kN
[+0.00]  d=21  1.Qxh2+ Qxh2 2.Re1 c4 3.Ke4 h4 4.Rc1 h3 5.Rb1 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Rb7 Bf2 9.Rh7 g1Q 10.Nxg1 Kxg1 11.c4 h1Q 12.Rxh1+ Kxh1 13.Kd5 Ba7 14.Kc6 Kg1 (0:00:34) 1653kN
[+0.14]  d=20  1.Qxh2+ Qxh2 2.Re1 c4 3.Ke4 h4 4.Rc1 h3 5.Rb1 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Rb7 Bf2 9.Rh7 g1Q 10.Nxg1 Bxg1 11.Rh8 Kg2 12.Kd5 Be3 13.Ke4 (0:00:10) 753kN
[+0.14]  d=19  1.Qxh2+ Qxh2 2.Re1 c4 3.Ke4 h4 4.Rc1 h3 5.Rb1 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Rb7 Bf2 9.Rh7 g1Q 10.Nxg1 Bxg1 11.Rh8 Kg2 12.Kd5 Be3 13.Ke4 (0:00:09) 588kN
[+0.14]  d=18  1.Qxh2+ Qxh2 2.Re1 c4 3.Ke4 h4 4.Rc1 h3 5.Ra1 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Ra8 Bb6 9.Rh8 g1Q 10.Nxg1 Kxg1 11.c4 h1Q 12.Rxh1+ Kxh1 13.Kd5 Kg2 (0:00:05) 348kN
[+0.14]  d=17  1.Qxh2+ Qxh2 2.Re1 c4 3.Ke4 h4 4.Rc1 h3 5.Ra1 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Ra8 Bb6 9.Rh8 g1Q 10.Nxg1 Kxg1 11.c4 h1Q 12.Rxh1+ Kxh1 13.Kd5 Kg2 (0:00:05) 285kN
[+0.03]  d=16  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Rd1 Bb6 10.Ke4 Kg3 11.Rd6 g1Q 12.Rg6+ Kh2 (0:00:02) 159kN
[+0.03]  d=15  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Rd1 Bb6 10.Ke4 Kg3 11.Rd6 g1Q 12.Rg6+ Kh2 (0:00:01) 109kN
[+0.14]  d=14  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Rb1 Bf2 10.c4 g1Q 11.Rxg1 Kxg1 12.Kd5 Kg2 (0:00:01) 74kN
[+0.14]  d=13  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Rb1 Bf2 10.c4 g1Q 11.Rxg1 Kxg1 12.Kd5 Kg2 (0:00:01) 39kN
[+0.14]  d=12  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Rb1 Bf2 10.c4 g1Q 11.Rxg1 Kxg1 12.Kd5 Kg2 (0:00:00) 23kN
[+0.24]  d=11  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Ke4 Bf2 10.Kf3 g1Q 11.Rxg1 Bxg1 12.c4 Bb6 13.Kf4 (0:00:00) 20kN
[+0.24]  d=10  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Ke4 (0:00:00) 16kN
[+0.20]  d=9  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 Qe5+ 7.Kxe5 h2 8.Nxh2 Kxh2 9.Ke4 (0:00:00) 16kN
[+0.20]  d=8  1.Qxh2+ Qxh2 2.Re1 h4 3.Ke4 h3 4.Rc1 c4 5.Kf5 c3 6.bxc3 (0:00:00) 16kN
- By Omega-three (**) Date 2009-08-13 07:19
Hi ALKa

This position may be suffering from the egtb bug. I say that because my quad after 10 minutes or so still shows a sore of 0 so I changed the hash and within a minute or so was showing 3.23 and now after 3.23 mins has 6.43 score.

for egtb bug see here : http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=11872

Regards

Mike
Up Topic Rybka Support & Discussion / Rybka Discussion / Deep rybka cannot solve a position I solved in 10 min. WTF??

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill