Rybka don't find mat in 4?

Dann Corbit · Post by **Dann Corbit** » Tue Aug 12, 2008 7:44 am

Jim Logan wrote:Interesting yes ... I can always solve as above using Single Processor version, but not with Multiprocessor version; at least not using multi-core. Have not tried setting the MC version to using only 1 core.

Dynamic solved with 4 threads, the others found mate, but it took a long time and they never found #4

Code: Select all

Analysis from Q&#58;\epd\sub\qc7.epd   
8/11/2008 10&#58;16&#58;55 PM Level&#58; 720 Seconds
Analyzing engine&#58; Rybka 3 Dynamic

1&#41; Qc7; c0 "Mate in 4"; 
    Searching move&#58; Qe5-c7
    Best move &#40;Rybka 3 Dynamic&#41;&#58; Qe5-c7
    identical moves! Found in&#58; 00&#58;00
      2	00&#58;00	          62	63.488	+3.63	Qe5d4+
      2	00&#58;00	          67	68.608	+4.06	Qe5d5
      3	00&#58;00	          73	74.752	+4.06	Qe5d5
      4	00&#58;00	          84	86.016	+4.06	Qe5d5
      5	00&#58;00	         125	128.000	+4.07	Qe5d5 Kb4a4
      6	00&#58;00	       1.552	1.589.248	+4.13	Qe5d5 Kb4a4 Qd5d6
      7	00&#58;00	       3.108	3.182.592	+4.30	Qe5d5 Kb4a4 Qd5d6 Ka4a5 Qd6a3+
      7+	00&#58;00	       5.333	170.656	+4.49	Qe5c7
      7	00&#58;00	       5.364	171.648	+4.49	Qe5c7 Kb4a4
      8	00&#58;00	       5.419	173.408	+4.49	Qe5c7 Kb4a4 Qc7d6 Ka4a5 Qd6a3+ Ka5b6 Qa3xb3 Kb6c5 Qb3c3+ Kc5b6 Qc3b4
      9	00&#58;00	       5.522	176.704	+4.49	Qe5c7 Kb4a4 Qc7d6 Ka4a5 Qd6a3+ Ka5b6 Qa3xb3 Kb6c5 Qb3c3+ Kc5b6 Qc3b4
     10	00&#58;00	       5.857	95.199	+4.50	Qe5c7 Kb4a4
     11	00&#58;00	       5.954	96.776	+4.50	Qe5c7 Kb4a4 Qc7d6 Ka4a5 Qd6a3+ Ka5b6 Qa3xb3 Kb6c5 Qb3c3+ Kc5b6 Qc3b4 Kb6a6 Qb4b3
     12	00&#58;00	       6.197	100.725	+4.50	Qe5c7 Kb4a4 Qc7d6 Ka4a5 Qd6a3+ Ka5b6 Qa3xb3 Kb6c5 Qb3c3+ Kc5b6 Qc3b4 Kb6a6 Qb4b3
     13	00&#58;00	       9.055	117.371	+4.50	Qe5c7 Kb4a4 Qc7d6 Ka4a5 Qd6a3+ Ka5b6 Qa3xb3 Kb6c5 Qb3c3+ Kc5b6 Qc3b4 Kb6a6 Qb4b3 Ka6a5 Qb3a3+
     14+	00&#58;00	      17.835	194.287	+5.11	Qe5c7
     14	00&#58;00	      17.957	195.616	+M4	Qe5c7
     14	00&#58;31	  28.272.555	923.745	+M4	Qe5c7
     15	03&#58;10	 177.687.866	955.476	+M4	Qe5c7
   8/11/2008 10&#58;41&#58;15 PM, Time for this analysis&#58; 00&#58;12&#58;00, Rated time&#58; 00&#58;00

1 of 1 matching moves
8/11/2008 10&#58;41&#58;16 PM, Total time&#58; 12&#58;24&#58;20 AM
Rated time&#58; 00&#58;00 = 0 Seconds

bob · Post by **bob** » Tue Aug 12, 2008 8:56 am

jpqy wrote:The most other chess engines find this mat in 4 around 00:00!!

How about Rybka 3?

MAT in 4!!

8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 7

JP.

I ran it to see what is going on...

Code: Select all

               12->   0.12   3.84   1. Qd4+ Ka5 2. Qc5 Ka4 3. Qd6 Ka5 4.
                                    Qa3+ Kb6 5. Qxb3 Kc5 6. Qc3+ Kb6 7.
                                    Qe3+ Ka5 <HT> &#40;s=3&#41;
               13     0.12     +1   1. Qd4+!!  &#40;1.7Mnps&#41;             
               13     0.17   4.40   1. Qd4+ Ka5 2. Qd2+ Ka6 3. Qd8 Kb7
                                    4. Qf6 Kc7 5. Qa6 b4 6. Qa5+ Kd6 7.
                                    Qxb4+ Kd5 8. Qa4 Kc5 9. Qxb3 &#40;s=2&#41;
               13     0.19     +1   1. Qc7!!   &#40;2.8Mnps&#41;             
               13     0.19     +3   1. Qc7!!   &#40;2.8Mnps&#41;             
               13     0.19     +M   1. Qc7!!   &#40;2.8Mnps&#41;             
               13     0.19  Mat04   1. Qc7 Ka4 2. Qc3 b4 3. Qe5 Ka3 4.
                                    Qa5#
               13->   0.20  Mat04   1. Qc7 Ka4 2. Qc3 b4 3. Qe5 Ka3 4.
                                    Qa5# &#40;s=5&#41;

Any time you introduce a position with zugzwang, you are inviting a program that uses null-move into a quagmire. This is just another one of those positions.

George Tsavdaris · Post by **George Tsavdaris** » Tue Aug 12, 2008 9:21 am

Albert Silver wrote:
jpqy wrote:The most other chess engines find this mat in 4 around 00:00!!

How about Rybka 3?

MAT in 4!!

8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 7

JP.
Takes a few seconds, but it finds it:

Analysis by Rybka 3 1-cpu 32-bit:

1.Qd4+ Ka5 2.Qe4 b4 3.Qd3
+- (5.02) Depth: 6 00:00:00 1kN
1.Qd4+ Ka5 2.Qe4 b4 3.Qd3 Kb6
+- (5.07) Depth: 7 00:00:00 2kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 8 00:00:00 12kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 9 00:00:00 13kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 10 00:00:00 14kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[]
+- (5.03) Depth: 11 00:00:00 15kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+
+- (5.03) Depth: 12 00:00:00 18kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+ Ka5[] 7.Qd6
+- (5.03) Depth: 13 00:00:00 22kN
1.Qc7
+- (#4) Depth: 13 00:00:27 2678kN

I wonder if you are using a different Rybka 3 32-bit SP than myself.

My Rybka 3 SP 32-bit 1 CPU can't find the move in any way.
128, 256, 512 MB hash tables, tablebases or non-tablebases, it doesn't matter. It just can't find it!

And we have i think the exact same hardware. What kind of Rybka 3 you are using? And i wonder if it's stronger than the commercial one....

For example:

Analysis by Rybka 3 1-cpu 32-bit 256 MB hash with TBs:

1.Qe5-d4+
+- (4.86) Depth: 2 00:00:00
1.Qe5-d4+
+- (5.06) Depth: 3 00:00:00
1.Qe5-d4+
+- (4.86) Depth: 3 00:00:00
1.Qe5-d4+
+- (5.03) Depth: 4 00:00:00
1.Qe5-d4+
+- (5.16) Depth: 5 00:00:00
1.Qe5-d4+
+- (5.12) Depth: 6 00:00:01 2kN
1.Qe5-d4+
+- (5.12) Depth: 7 00:00:01 7kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-e3+ Kb6-c6
+- (5.09) Depth: 8 00:00:01 9kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-e3+ Kb6-c6 4.Qe3xb3 Kc6-c5
+- (5.09) Depth: 9 00:00:01 11kN
1.Qe5-d4+
+- (5.10) Depth: 10 00:00:01 12kN
1.Qe5-d4+
+- (5.79) Depth: 11 00:00:02 135kN, tb=2
1.Qe5-d4+
+- (5.86) Depth: 12 00:00:05 422kN, tb=3
1.Qe5-d4+
+- (5.98) Depth: 13 00:01:14 7852kN, tb=10
1.Qe5-d4+
+- (6.91) Depth: 14 00:01:36 10458kN, tb=52
1.Qe5-d4+
+- (#27) Depth: 15 00:01:43 11269kN, tb=123
1.Qe5-d4+ Kb4-a5
+- (#27) Depth: 15 00:01:44 11284kN, tb=124

Also i see in your analysis there are the [] brackets. What GUI are you using? The new one from Chessbase that comes with Rybka 3?
If so perhaps this is the reason that your Rybka 3 32-bit finds it.
If so can you try Rybka 3 in an older Chessbase GUI?

Uri Blass · Post by **Uri Blass** » Tue Aug 12, 2008 10:31 am

George Tsavdaris wrote:
Albert Silver wrote:
jpqy wrote:The most other chess engines find this mat in 4 around 00:00!!

How about Rybka 3?

MAT in 4!!

8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 7

JP.
Takes a few seconds, but it finds it:

Analysis by Rybka 3 1-cpu 32-bit:

1.Qd4+ Ka5 2.Qe4 b4 3.Qd3
+- (5.02) Depth: 6 00:00:00 1kN
1.Qd4+ Ka5 2.Qe4 b4 3.Qd3 Kb6
+- (5.07) Depth: 7 00:00:00 2kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 8 00:00:00 12kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 9 00:00:00 13kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+
+- (5.03) Depth: 10 00:00:00 14kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[]
+- (5.03) Depth: 11 00:00:00 15kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+
+- (5.03) Depth: 12 00:00:00 18kN
1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+ Ka5[] 7.Qd6
+- (5.03) Depth: 13 00:00:00 22kN
1.Qc7
+- (#4) Depth: 13 00:00:27 2678kN
I wonder if you are using a different Rybka 3 32-bit SP than myself.

My Rybka 3 SP 32-bit 1 CPU can't find the move in any way.
128, 256, 512 MB hash tables, tablebases or non-tablebases, it doesn't matter. It just can't find it!

And we have i think the exact same hardware. What kind of Rybka 3 you are using? And i wonder if it's stronger than the commercial one....

For example:

Analysis by Rybka 3 1-cpu 32-bit 256 MB hash with TBs:

1.Qe5-d4+
+- (4.86) Depth: 2 00:00:00
1.Qe5-d4+
+- (5.06) Depth: 3 00:00:00
1.Qe5-d4+
+- (4.86) Depth: 3 00:00:00
1.Qe5-d4+
+- (5.03) Depth: 4 00:00:00
1.Qe5-d4+
+- (5.16) Depth: 5 00:00:00
1.Qe5-d4+
+- (5.12) Depth: 6 00:00:01 2kN
1.Qe5-d4+
+- (5.12) Depth: 7 00:00:01 7kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-e3+ Kb6-c6
+- (5.09) Depth: 8 00:00:01 9kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-e3+ Kb6-c6 4.Qe3xb3 Kc6-c5
+- (5.09) Depth: 9 00:00:01 11kN
1.Qe5-d4+
+- (5.10) Depth: 10 00:00:01 12kN
1.Qe5-d4+
+- (5.79) Depth: 11 00:00:02 135kN, tb=2
1.Qe5-d4+
+- (5.86) Depth: 12 00:00:05 422kN, tb=3
1.Qe5-d4+
+- (5.98) Depth: 13 00:01:14 7852kN, tb=10
1.Qe5-d4+
+- (6.91) Depth: 14 00:01:36 10458kN, tb=52
1.Qe5-d4+
+- (#27) Depth: 15 00:01:43 11269kN, tb=123
1.Qe5-d4+ Kb4-a5
+- (#27) Depth: 15 00:01:44 11284kN, tb=124

Also i see in your analysis there are the [] brackets. What GUI are you using? The new one from Chessbase that comes with Rybka 3?
If so perhaps this is the reason that your Rybka 3 32-bit finds it.
If so can you try Rybka 3 in an older Chessbase GUI?

My rybka3 under Fritz8 gui could clearly find it with no tablebases and
512 mbytes hash.

I could reproduce it

Strelka 2.10 - so k
8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 1

Analysis by Rybka 3 1-cpu 32-bit :

1.Qe5-d4+
+- (4.78) Depth: 2 00:00:00
1.Qe5-d4+
+- (4.78) Depth: 3 00:00:00
1.Qe5-d4+
+- (4.84) Depth: 4 00:00:00
1.Qe5-d4+ Kb4-a5 2.Qd4-e4
+- (5.02) Depth: 5 00:00:00
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 b5-b4 3.Qe4-d3
+- (5.02) Depth: 6 00:00:00 1kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 b5-b4 3.Qe4-d3 Ka5-b6
+- (5.07) Depth: 7 00:00:00 2kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+
+- (5.03) Depth: 8 00:00:00 12kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+
+- (5.03) Depth: 9 00:00:00 13kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+
+- (5.03) Depth: 10 00:00:00 14kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+ Kc5-b6
+- (5.03) Depth: 11 00:00:01 15kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+ Kc5-b6 6.Qc3-d4+
+- (5.03) Depth: 12 00:00:01 18kN
1.Qe5-d4+ Kb4-a5 2.Qd4-e4 Ka5-b6 3.Qe4-b4 Kb6-c6 4.Qb4xb3 Kc6-c5 5.Qb3-c3+ Kc5-b6 6.Qc3-d4+ Kb6-a5 7.Qd4-d6
+- (5.03) Depth: 13 00:00:01 22kN
1.Qe5-c7
+- (#4) Depth: 13 00:00:26 2678kN

(so k, 12.08.2008)

Uri

George Tsavdaris · Post by **George Tsavdaris** » Tue Aug 12, 2008 11:51 am

Uri Blass wrote:
George Tsavdaris wrote:
Albert Silver wrote:
jpqy wrote:The most other chess engines find this mat in 4 around 00:00!!

How about Rybka 3?

MAT in 4!!

8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 7

JP.
Takes a few seconds, but it finds it:

Analysis by Rybka 3 1-cpu 32-bit:

1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+ Ka5[] 7.Qd6
+- (5.03) Depth: 13 00:00:00 22kN
1.Qc7
+- (#4) Depth: 13 00:00:27 2678kN
I wonder if you are using a different Rybka 3 32-bit SP than myself.

My Rybka 3 SP 32-bit 1 CPU can't find the move in any way.
128, 256, 512 MB hash tables, tablebases or non-tablebases, it doesn't matter. It just can't find it!

And we have i think the exact same hardware. What kind of Rybka 3 you are using? And i wonder if it's stronger than the commercial one....

For example:

Analysis by Rybka 3 1-cpu 32-bit 256 MB hash with TBs:

1.Qe5-d4+
+- (6.91) Depth: 14 00:01:36 10458kN, tb=52
1.Qe5-d4+
+- (#27) Depth: 15 00:01:43 11269kN, tb=123
1.Qe5-d4+ Kb4-a5
+- (#27) Depth: 15 00:01:44 11284kN, tb=124

Also i see in your analysis there are the [] brackets. What GUI are you using? The new one from Chessbase that comes with Rybka 3?
If so perhaps this is the reason that your Rybka 3 32-bit finds it.
If so can you try Rybka 3 in an older Chessbase GUI?
My rybka3 under Fritz8 gui could clearly find it with no tablebases and
512 mbytes hash.

I could reproduce it

Analysis by Rybka 3 1-cpu 32-bit :

1.Qe5-c7
+- (#4) Depth: 13 00:00:26 2678kN

I think it has something to do with the contempt=15 value.

I have reproduced it easily now that i have chosen to do analysis with Rybka in multi-variation mode with 1 line of analysis.
In analysis Rybka 3 has contempt=0.

In previous times the analysis of Rybka i gave wasn't with the analysis button that means i didn't press the analysis button but was choosing fixed time(3600 seconds) and let Rybka think and play.
So Rybka was like playing a game but just with 1 hour thought. That way it was using contempt=15.

The analysis you, i and Albert have provided can come up with 2 ways in Chessbase GUI:
1)Pressing spacebar button and Rybka will analyze according to the time it has(a big time as i choose is suitable for analysis) in single line PVs.
2)Pressing the analysis button and choosing with the + - buttons, one single line each time to appear.

So i hope you can confirm you did multi-variation analysis with choosing one single line to appear.
In a different case i would be very puzzled again.

Uri Blass · Post by **Uri Blass** » Tue Aug 12, 2008 1:28 pm

George Tsavdaris wrote:
Uri Blass wrote:
George Tsavdaris wrote:
Albert Silver wrote:
jpqy wrote:The most other chess engines find this mat in 4 around 00:00!!

How about Rybka 3?

MAT in 4!!

8/8/8/1p2Q3/1k6/1p6/1p6/bK6 w - - 0 7

JP.
Takes a few seconds, but it finds it:

Analysis by Rybka 3 1-cpu 32-bit:

1.Qd4+ Ka5[] 2.Qe4 Kb6 3.Qb4 Kc6[] 4.Qxb3 Kc5 5.Qc3+ Kb6[] 6.Qd4+ Ka5[] 7.Qd6
+- (5.03) Depth: 13 00:00:00 22kN
1.Qc7
+- (#4) Depth: 13 00:00:27 2678kN
I wonder if you are using a different Rybka 3 32-bit SP than myself.

My Rybka 3 SP 32-bit 1 CPU can't find the move in any way.
128, 256, 512 MB hash tables, tablebases or non-tablebases, it doesn't matter. It just can't find it!

And we have i think the exact same hardware. What kind of Rybka 3 you are using? And i wonder if it's stronger than the commercial one....

For example:

Analysis by Rybka 3 1-cpu 32-bit 256 MB hash with TBs:

1.Qe5-d4+
+- (6.91) Depth: 14 00:01:36 10458kN, tb=52
1.Qe5-d4+
+- (#27) Depth: 15 00:01:43 11269kN, tb=123
1.Qe5-d4+ Kb4-a5
+- (#27) Depth: 15 00:01:44 11284kN, tb=124

Also i see in your analysis there are the [] brackets. What GUI are you using? The new one from Chessbase that comes with Rybka 3?
If so perhaps this is the reason that your Rybka 3 32-bit finds it.
If so can you try Rybka 3 in an older Chessbase GUI?
My rybka3 under Fritz8 gui could clearly find it with no tablebases and
512 mbytes hash.

I could reproduce it

Analysis by Rybka 3 1-cpu 32-bit :

1.Qe5-c7
+- (#4) Depth: 13 00:00:26 2678kN
I think it has something to do with the contempt=15 value.

I have reproduced it easily now that i have chosen to do analysis with Rybka in multi-variation mode with 1 line of analysis.
In analysis Rybka 3 has contempt=0.

In previous times the analysis of Rybka i gave wasn't with the analysis button that means i didn't press the analysis button but was choosing fixed time(3600 seconds) and let Rybka think and play.
So Rybka was like playing a game but just with 1 hour thought. That way it was using contempt=15.

The analysis you, i and Albert have provided can come up with 2 ways in Chessbase GUI:
1)Pressing spacebar button and Rybka will analyze according to the time it has(a big time as i choose is suitable for analysis) in single line PVs.
2)Pressing the analysis button and choosing with the + - buttons, one single line each time to appear.

So i hope you can confirm you did multi-variation analysis with choosing one single line to appear.
In a different case i would be very puzzled again.

I can confirm it.
It is the usual way that I make analysis.
contempt=15 is not objective and I prefer to have objective analysis

Uri

Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?

Re: Rybka don't find mat in 4?