Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A move of Junior7 that I could not reproduce at long time control.

Author: Brian Kostick

Date: 16:00:22 07/11/01

Go up one level in this thread


On July 11, 2001 at 07:57:19, Chessfun wrote:

>On July 11, 2001 at 07:54:14, Harald Faber wrote:
>
>>On July 11, 2001 at 07:50:11, Chessfun wrote:
>>
>>>On July 11, 2001 at 07:42:49, Uri Blass wrote:
>>>
>>>>On July 11, 2001 at 06:26:29, Harald Faber wrote:
>>>>
>>>>>On July 11, 2001 at 05:52:37, Uri Blass wrote:
>>>>>
>>>>>>
>>>>>>This position is from haraled faver's games.
>>>>>
>>>>>Ähm, I don't care much if my name is not written correctly, but this spelling is
>>>>>totally wrong. :-)
>>>>>It is Harald Faber, easy, isn't it? ;-)
>>>>
>>>>yes
>>>>
>>>>I will remember in the future not to put e between the 'l' and the 'd'.
>>>>
>>>>I also found that the bug is in Junior and not in the autoplayer(see the end of
>>>>the post for proof.
>>>>>
>>>>>>Junior 7,K - ChessTiger 14,K
>>>>>>[D]r1r2k2/pb3pp1/4p2p/3nP3/Nq6/3Q1N2/1P3PPP/R2R2K1 w - - 0 1
>>>>>>
>>>>>>Junior played here 22.Nc3
>>>>>>
>>>>>>I gave my Junior7 to anlayze the same position and it did not like Nc3
>>>>>
>>>>>
>>>>>What hardware do you use?
>>>>
>>>>PIII800.
>>>>>
>>>>>
>>>>>>Analysis by Junior 7:
>>>>>>
>>>>>>  =  (-0.02)   Depth: 9   00:00:01  694kN
>>>>>>22.Qh7 Nf4 23.Rd7 Ke8 24.Rdd1 Ke7 25.b3 Nh3+ 26.gxh3 Bxf3 27.Rd6 Kf8
>>>>>>  =  (-0.05)   Depth: 12   00:00:11  8344kN
>>>>>>22.Qh7 Nf4 23.Rd7 Ng6 24.Rd4 Qb5 25.Nh4 Nxh4 26.Rxh4 Qxe5 27.Rb4 Bd5 28.Nc3 a5
>>>>>>  =  (-0.25)   Depth: 14   00:01:06  46907kN
>>>>>>22.h3 Nf4 23.Qd6+ Qxd6 24.exd6 Bxf3 25.gxf3 Rc2 26.h4 Rd8 27.Nc3 Nh3+ 28.Kg2
>>>>>>Nxf2
>>>>>>  =  (-0.24)   Depth: 14   00:01:10  49142kN
>>>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4 25.Rdc1 Rab8 26.Rxc8+
>>>>>>  =  (-0.05)   Depth: 14   00:01:14  52426kN
>>>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4 25.Rdc1 Nc3 26.Nd4
>>>>>>  =  (-0.16)   Depth: 15   00:01:24  58639kN
>>>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4 25.Rdc1 Nc3 26.Nd4
>>>>>>  =  (-0.16)   Depth: 16   00:02:36  105349kN
>>>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4 25.Rdc1 Nc3 26.Nd4 Be4
>>>>>>  =  (-0.17)   Depth: 17   00:05:12  203618kN
>>>>>>
>>>>>>(Blass, Tel-aviv 11.07.2001)
>>>>>>
>>>>>>I tried to give it the same position at tournament condition and here is the
>>>>>>analysis
>>>>>
>>>>>
>>>>>What do you mean by tournament condition?
>>>>
>>>>I clicked on games,level,long game and chose 2 hours/40 moves.
>>>>
>>>>I later clicked on game and move now so junior started to calculate.
>>>>
>>>>Junior had all the game so it practically had 2 hours for less than 40 moves but
>>>>I could reproduce the same result in 1 hour/40 moves.
>>>>
>>>>
>>>>>
>>>>>
>>>>>>Junior 7 - Blass,U
>>>>>>r1r2k2/pb3pp1/4p2p/3nP3/Nq6/3Q1N2/1P3PPP/R2R2K1 w - - 0 1
>>>>>>
>>>>>>Analysis by Junior 7:
>>>>>>
>>>>>>  =  (-0.02)   Depth: 9   00:00:01  638kN
>>>>>>22.Qh7 Nf4 23.Rd7 Ke8 24.Rdd1
>>>>>>  =  (0.00)   Depth: 12   00:00:12  9070kN
>>>>>>22.Nc3 Nxc3 23.bxc3 Qxc3 24.Qd6+ Kg8 25.Nd4 Rd8 26.Rac1 Qxc1 27.Rxc1 Rxd6
>>>>>>28.exd6
>>>>>>  =  (-0.03)   Depth: 12   00:00:14  10323kN
>>>>>>22.Nc3 Qxb2 23.Ne4 Nf4 24.Qd6+ Kg8 25.Rab1 Qe2 26.Rxb7 Qxe4 27.Qe7 Nh3+ 28.Kf1
>>>>>>Qc4+ 29.Ke1 Qc3+ 30.Kf1
>>>>>>  =  (0.00)   Depth: 14   00:00:21  14864kN
>>>>>>22.Qa3 Qxa3 23.Rxa3 Ne7 24.Nd4 Ng6 25.Re3 Rab8 26.Nc3
>>>>>>  =  (-0.14)   Depth: 14   00:00:52  35689kN
>>>>>>22.Qa3 Qxa3 23.Rxa3 Nf4 24.Ne1 Ne2+ 25.Kh1 Rab8 26.Nc3 Nxc3 27.bxc3 a5
>>>>>>  =  (-0.17)   Depth: 16   00:01:02  42938kN
>>>>>>22.b3 Kg8 23.Qd2 a5 24.Qxb4 axb4 25.Rdc1 Nc3 26.Nd4
>>>>>>  =  (-0.16)   Depth: 16   00:01:39  67309kN
>>>>>>
>>>>>>(Blass, Tel-aviv 11.07.2001)
>>>>>>
>>>>>>
>>>>>>Here is the analysis of Junior after 22.Nc3
>>>>>>
>>>>>>Junior 7,K - ChessTiger 14,K
>>>>>>r1r2k2/pb3pp1/4p2p/3nP3/1q6/2NQ1N2/1P3PPP/R2R2K1 b - - 0 1
>>>>>>
>>>>>>Analysis by Junior 7:
>>>>>>
>>>>>>22...Nxc3 23.bxc3 Rxc3 24.Qh7 Bxf3
>>>>>>  ³  (-0.42)   Depth: 3   00:00:00
>>>>>>22...Nxc3 23.bxc3 Qxc3 24.Qxc3
>>>>>>  =  (-0.17)   Depth: 6   00:00:00  3kN
>>>>>>22...Nf4 23.Qh7 Nxg2 24.Ra4 Qxb2 25.Rb1 Qxc3 26.Qh8+ Ke7 27.Rxb7+
>>>>>>  ³  (-0.29)   Depth: 6   00:00:00  14kN
>>>>>>22...Nf4 23.Qd6+ Qxd6 24.exd6 Bxf3 25.gxf3 Rc4 26.d7 Ke7
>>>>>>  ²  (0.36)   Depth: 9   00:00:00  172kN
>>>>>>22...Nxc3 23.bxc3 Qxc3 24.Qd6+
>>>>>>  =  (-0.17)   Depth: 9   00:00:00  184kN
>>>>>>22...Nxc3 23.bxc3 Qxc3 24.Qd6+ Kg8 25.Nd4 a5 26.Qd7 Rc7 27.Ra3 Qe1+ 28.Rxe1 Rxd7
>>>>>>  ³  (-0.36)   Depth: 12   00:00:03  1590kN
>>>>>>22...Nxc3 23.bxc3 Qxc3 24.Qh7 Bxf3 25.gxf3 a5 26.f4 Qf3 27.Rd3 Qe2 28.Rd7
>>>>>>  ³  (-0.43)   Depth: 15   00:00:29  17321kN
>>>>>>22...Nxc3 23.bxc3 Qxc3 24.Qd6+ Kg8 25.Nd4 Rd8 26.Qa3 Qxa3 27.Rxa3 a5 28.f4 a4
>>>>>>29.Kf2 Be4
>>>>>>  ³  (-0.59)   Depth: 17   00:02:35  89481kN
>>>>>>
>>>>>>(Blass, Tel-aviv 11.07.2001)
>>>>>>
>>>>>>I wonder if the reason for Nc3 was some learning from hash tables,different size
>>>>>>of hash tables, autoplayer bug or another reason(Junior used 848 seconds to
>>>>>>caluclate the move Nc3 based on the pgn)
>>>>>>
>>>>>>Uri
>>>>>
>>>>>1) Note that the first move calculated by Junior in this game (no. 5) is move
>>>>>#17. Both programs played from the book up to then. So at move #22 there is some
>>>>>moves hash inside.
>>>>>2) I used 128MB hash, which hash size did you try?
>>>>
>>>>64 Mbytes
>>>>
>>>>>3) Not Junior took 848 seconds, but Tiger took 848 sec for 21...Kf8. Sure,
>>>>>Junior moved immediately after this 848sec but Juniors move #21 was also
>>>>>immediately so I'd say all this hash/permanent brain may cause the difference.
>>>>
>>>>I understand it but Junior calculated for 848 seconds when it was tiger's move
>>>>so my statement that Junior used 848 seconds is correct.
>>>>
>>>>I agree that more hash and permanent brain can cause difference but they are
>>>>supposed to be productive and not counter productive.
>>>>
>>>>Junior has no problem to avoid Nc3 at depth 14.
>>>>
>>>>I decided to try it at game condition and was surprised to find that Junior can
>>>>play it at depth 17 with my PIII800.
>>>>
>>>>I gave it to play 21.Rfd1 at 40 moves/1 hour and let it ponder for a long time.
>>>>
>>>>This is the surprising result:
>>>>
>>>>Junior never considered Nc3 as a good move but suddenly at depth 17 found this
>>>>move.
>>>>
>>>>Junior 7 - Blass,U
>>>>r1r5/pb2kpp1/4p2p/3nP3/Nq6/3Q1N2/1P3PPP/R2R2K1 b - - 0 1
>>>>
>>>>Analysis by Junior 7:
>>>>
>>>>22.b3 Rd8
>>>>  =  (0.07)   Depth: 3   00:00:00
>>>>22.h3 Rd8
>>>>  =  (0.10)   Depth: 3   00:00:00
>>>>22.Nd4 Rd8
>>>>  =  (0.13)   Depth: 3   00:00:00
>>>>22.Qh7 Rd8
>>>>  ²  (0.27)   Depth: 3   00:00:00
>>>>22.Qh7 Nf4
>>>>  =  (0.00)   Depth: 6   00:00:00  1kN
>>>>22.Qa3 Qxa3
>>>>  =  (-0.12)   Depth: 6   00:00:00  1kN
>>>>22.Qa3 Qxa3
>>>>  =  (-0.12)   Depth: 9   00:00:00  3kN
>>>>22.Qa3 Qxa3
>>>>  =  (-0.12)   Depth: 12   00:00:00  30kN
>>>>22.Qa3 Qxa3 23.Rxa3 Ne7 24.Nd4 Ng6 25.Re3 Rab8 26.Nc3
>>>>  =  (-0.14)   Depth: 14   00:00:02  830kN
>>>>22.Qa3 Qxa3 23.Rxa3 Nf4 24.Ne1 Ne2+ 25.Kh1 Rab8 26.Nc3 Nxc3 27.bxc3 a5
>>>>  =  (-0.17)   Depth: 16   00:00:47  30491kN
>>>>22.b3 Kg8 23.Qd2 a5 24.Qxb4 axb4 25.Rdc1 Nc3 26.Nd4
>>>>  =  (-0.16)   Depth: 16   00:04:33  178849kN
>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4
>>>>  =  (-0.17)   Depth: 17   00:06:21  245439kN
>>>>22.Nc3 Nxc3 23.bxc3 Qxc3 24.Qh7 Bxf3 25.gxf3 Qxf3 26.Rd7 Qg4+ 27.Kf1 Qh3+ 28.Ke2
>>>>Qg4+
>>>>  =  (0.00)   Depth: 17   00:08:00  305132kN
>>>>22.Nc3 Nxc3 23.bxc3 Qxc3 24.Qh7 Bxf3 25.gxf3 a5 26.f4 Qf3 27.Rd3 Qe2 28.Rd7
>>>>  ³  (-0.43)   Depth: 18   00:15:15  574325kN
>>>>22.b3 Kg8 23.Qd4 a5 24.Qxb4 axb4 25.Nd4 Kf8 26.g4
>>>>  =  (-0.18)   Depth: 18   00:16:27  618889kN
>>>>
>>>>(Blass, Tel-aviv 11.07.2001)
>>>>
>>>>
>>>>I guess that similiar analysis convinced Junior in your game except the fact
>>>>that tiger did not use enough time to let Junior to get into depth 18.
>>>
>>>It is still rather odd.
>>>
>>>Here is what I get TB 945 w/128 mb hash
>>>New position
>>>r1r2k2/pb3pp1/4p2p/3nP3/Nq6/3Q1N2/1P3PPP/R2R2K1 w - - 0 1
>>>
>>>Analysis by Junior 7:
>>>1.Qh7 Nf4 2.Rd7 Ng6 3.Rd4 Qb5 4.Nh4 Nxh4 5.Rxh4 Qxe5 6.Rb4 Bd5 7.Nc3 a5
>>>  =  (-0.25)   Depth: 14   00:01:00  46973kN
>>>1.h3 Nf4 2.Qd6+ Qxd6 3.exd6 Bxf3 4.gxf3 Rc2 5.h4 Rd8 6.Nc3 Nh3+ 7.Kg2 Nxf2
>>>  =  (-0.24)   Depth: 14   00:01:03  49207kN
>>>1.b3 Kg8 2.Qd4 a5 3.Qxb4 axb4 4.Rdc1 Rab8 5.Rxc8+
>>>  =  (-0.05)   Depth: 14   00:01:08  52495kN
>>>1.b3 Kg8 2.Qd4 a5 3.Qxb4 axb4 4.Rdc1 Nc3 5.Nd4
>>>  =  (-0.16)   Depth: 15   00:01:17  58710kN
>>>1.b3 Kg8 2.Qd4 a5 3.Qxb4 axb4 4.Rdc1 Nc3 5.Nd4
>>>  =  (-0.16)   Depth: 16   00:02:23  105308kN
>>>1.b3 Kg8 2.Qd4 a5 3.Qxb4 axb4 4.Rdc1 Nc3 5.Nd4 Be4
>>>  =  (-0.17)   Depth: 17   00:04:43  201433kN
>>>1.b3 Kg8 2.Qd4 a5 3.Qxb4 axb4 4.Nd4 Kf8 5.Rdc1 Ke7
>>>  =  (-0.12)   Depth: 18   00:11:44  502036kN
>>>
>>>Sarah.
>>
>>
>>This is not odd, look at Uri's post, you have a similar result.
>
>Maybe similar but I never see Nc3.
>
>Sarah.

Sarah and all good readers,

    I think this might be similar to:

http://www.icdchess.com/forums/1/message.shtml?178904

  But for sure I am but quessing! Still when I need best move from engine I tend
to set Game/Levels/Fixed Time...  to 99999 seconds and then press go button for
the engine to start thinking.   I tend to think that Infinite Analysis (Alt +
F2) does not bring out the best of engines. So much so that I consider it a bug,
but maybe I am not understanding the underling code?

  I have noted similar discrepancies with Junior 6 and Hiarcs 7.32. This was
back awhile when some of the forum was doing some 2 hour test suite. Sorry I
don't recall the test name.

  If you feel I am off track or not on topic I hope you will excuse this post.

Regards, Brian Kostick



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.