Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Shredder fans: results about S8 vs S9

Author: Henrik Dinesen

Date: 14:08:49 10/03/05

Go up one level in this thread


On October 03, 2005 at 11:52:22, Uri Blass wrote:

>On October 03, 2005 at 11:04:15, Henrik Dinesen wrote:
>
>>On October 02, 2005 at 11:17:56, Keith Hyams wrote:
>>
>>>I haven't got the sources for these comments - they are just fragments of
>>>memory.
>>>  The increase in Elo rating between S8 and S9 is small - not enough for me to
>>>consider the upgrade. In one set of results that I saw it S9 was +2. Most
>>>results tables put the increase as slightly greater and I have never seen S8
>>>come out higher than S9.
>>>
>>>One established correspondance player stated that he used Fritz 8 instead of
>>>Shredder because the "analysis of Shredder is all over the place". I was
>>>intrigued by this statement because I was not sure how an engine with analysis
>>>all over the place got the right answer so often. However I think it is
>>>universally accepted that Shredder analysis is optimistic.
>>>
>>>You may want to have a look at Fruit 2.1 if you are looking for consistency in
>>>output. Fruit2.1 is freeware and Fruit2.2 is commercial.
>>>
>>>                                  Keith
>>
>>Shredder 9 is more reliable than earlier versions in analysis, but I usually set
>>selectivety to zero if I want more trustworthy PVs when using Shredder.
>>
>>Regards
>>Henrik
>
>Based on experience shredder is not reliable for analysis and it is no problem
>of selectivity
>
>[Event "10'/40+10'/40+10'/40"]
>[Site "?"]
>[Date "2005.10.03"]
>[Round "?"]
>[White "-"]
>[Black "-"]
>[Result "*"]
>[PlyCount "31"]
>[TimeControl "40/600:40/600:40/600"]
>
>1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 dxc4 5. a4 Bf5 6. Nh4 Bg4 7. h3 Bd7 8. e3
>b5 9. Nf3 e6 10. Ne5 a6 11. Qf3 Ra7 12. Qg3 Rc7 13. Be2 b4 14. Nb1 c5 15. Nd2
>c3 16. bxc3 *
>
>
>Here is an example
>
>The evaluation before 15...c3 is positive for white
>evaluation after 15...c3 is positive for black and it is not question of time.
>
>If you delete learning you will find that evaluation before 15...c3 can be
>positive for black even after hours of search and if you delete learning you
>will find that evaluation after 15...c3 is positive for black even after a short
>time.
>
>
>- - -, 10'/40+10'/40+10'/40
>[D]1n1qkb1r/2rb1ppp/p3pn2/2p1N3/PppP4/4P1QP/1P1NBPP1/R1B1K2R b KQk - 0 1
>
>Analysis by Shredder 9 UCI:
>
>15...cxd4
>  ³  (-0.50)   Depth: 1/4   00:00:00
>15...cxd4 16.Bxc4
>  ³  (-0.67)   Depth: 1/6   00:00:00
>15...cxd4 16.Bxc4
>  ³  (-0.67)   Depth: 2/6   00:00:00
>15...cxd4 16.Bxc4
>  ³  (-0.42)   Depth: 3/8   00:00:00
>15...cxd4 16.Ndxc4 dxe3
>  ³  (-0.39)   Depth: 3/9   00:00:00
>15...cxd4 16.Ndxc4
>  =  (-0.05)   Depth: 4/9   00:00:00
>15...cxd4 16.Ndxc4 dxe3 17.0-0
>  =  (-0.05)   Depth: 4/9   00:00:00
>15...cxd4 16.Ndxc4
>  ²  (0.53)   Depth: 5/9   00:00:00
>15...cxd4 16.Ndxc4 dxe3 17.Bxe3
>  ²  (0.53)   Depth: 5/9   00:00:00
>15...cxd4 16.Ndxc4
>  ²  (0.28)   Depth: 6/10   00:00:00  8kN
>15...cxd4 16.Ndxc4 dxe3 17.Bxe3 Nc6 18.0-0 Nxe5
>  =  (0.17)   Depth: 6/12   00:00:00  10kN
>15...cxd4 16.Ndxc4
>  ²  (0.42)   Depth: 7/13   00:00:00  14kN
>15...cxd4 16.Ndxc4 Nc6
>  ²  (0.42)   Depth: 7/15   00:00:00  22kN
>15...cxd4 16.Ndxc4 Nc6 17.0-0 g6 18.Nxc6 Bxc6 19.exd4 Be7 20.Bf4 Rd7
>  ²  (0.53)   Depth: 8/18   00:00:01  51kN
>15...cxd4 16.Ndxc4 Nc6 17.0-0 g6 18.Nxc6 Bxc6 19.exd4 Be7 20.Bf4 Rd7
>  ²  (0.53)   Depth: 9/18   00:00:02  93kN
>15...cxd4 16.Ndxc4 Nc6 17.0-0 g6 18.Nxc6 Bxc6 19.exd4 Be7 20.Bf4 Ne4 21.Qe3
>  ²  (0.57)   Depth: 10/23   00:00:03  225kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5
>  ²  (0.53)   Depth: 10/23   00:00:03  368kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5
>  ²  (0.53)   Depth: 10/23   00:00:03  369kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5
>  ²  (0.53)   Depth: 11/23   00:00:04  390kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5
>  ²  (0.53)   Depth: 12/23   00:00:05  616kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5
>  ²  (0.53)   Depth: 13/23   00:00:06  1099kN
>15...c3 16.bxc3 bxc3 17.Nb3 cxd4 18.exd4 Nc6 19.0-0 Nxe5 20.dxe5 Ne4 21.Qf3 Nc5
>22.Nxc5 Bxc5
>  ²  (0.53)   Depth: 14/27   00:00:08  3468kN
>15...c3 16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 15/30   00:00:23  9636kN
>
>(,  03.10.2005)
>
>- - -, 10'/40+10'/40+10'/40
>1n1qkb1r/2rb1ppp/p3pn2/2p1N3/Pp1P4/2p1P1QP/1P1NBPP1/R1B1K2R w KQk - 0 1
>
>Analysis by Shredder 9 UCI:
>
>16.Nxd7
>  µ  (-1.03)   Depth: 1/4   00:00:00
>16.Nxd7
>  -+  (-1.95)   Depth: 1/5   00:00:00
>16.Nxd7 cxd2+ 17.Bxd2 Rxd7
>  -+  (-3.21)   Depth: 1/10   00:00:00
>16.0-0
>  -+  (-2.87)   Depth: 1/10   00:00:00
>16.0-0
>  -+  (-1.93)   Depth: 1/10   00:00:00
>16.0-0
>  -+  (-1.93)   Depth: 1/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 1/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 1/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 2/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 3/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 4/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 5/10   00:00:00
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 6/10   00:00:00  4kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 7/10   00:00:00  12kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 8/10   00:00:00  26kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 9/10   00:00:00  26kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 10/10   00:00:01  37kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 11/11   00:00:01  99kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 12/12   00:00:03  210kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 13/13   00:00:04  424kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.0-0 Ne4 19.Qf4 Bd5 20.Rb1
>  ²  (0.61)   Depth: 14/14   00:00:06  664kN
>16.bxc3 bxc3
>  ²  (0.36)   Depth: 15/31   00:00:08  2166kN
>16.bxc3 bxc3
>  =  (-0.14)   Depth: 15/31   00:00:10  4098kN
>16.bxc3 bxc3 17.Ndf3 Bc6 18.Nxc6 Nxc6 19.0-0 cxd4 20.Bxa6 Qd6 21.Qxd6 Bxd6
>22.exd4 h6 23.Re1 Nd5
>  ³  (-0.41)   Depth: 15/31   00:00:22  9426kN
>
>(,  03.10.2005)


Didn't say reliable, just _more_ .
And I agree. Good example. BTW, I don't know where to reset P-learning in
CB-version, except deleting it.

Referring to selectivity: Reducing it has shown less weird PV's of the kind with
piece-losses. On the other hand, I haven't spent 100's of hours analysing with
Shredder.

Regards
Henrik



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.