Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is Crafty broke?!

Author: Slater Wold

Date: 03:23:11 09/25/01

Go up one level in this thread


On September 25, 2001 at 03:07:18, Jouni Uski wrote:

>On September 25, 2001 at 01:29:09, Uri Blass wrote:
>
>>On September 24, 2001 at 20:33:21, Slater Wold wrote:
>>
>>>On September 24, 2001 at 19:35:51, Uri Blass wrote:
>>>
>>>>On September 24, 2001 at 19:10:43, Dann Corbit wrote:
>>>>
>>>>>There is no problem with crafty.
>>>>>It's a problem with the counters.
>>>>>
>>>>>The biggest number a long can hold is about 2 billion.
>>>>>
>>>>>That overflows very easily.
>>>>>
>>>>>If you let crafty run for more than a few minutes, the NPS figures have no
>>>>>meaning.  You used to be able to get negative values, but I think at least that
>>>>>has been fixed.
>>>>
>>>>Note that based on my experience of long analysis
>>>>Deep Fritz has not the same problem.
>>>>
>>>>Uri
>>>
>>>You're correct:
>>>
>>>W,S - Deep Junior 7
>>>rnbqk2r/1p3ppp/p7/1NpPp3/QPP1P1n1/P4N2/4KbPP/R1B2B1R b kq - 0 1
>>>
>>>Analysis by Deep Junior 7:
>>>
>>>1...cxb4 2.Nc7+ Kf8 3.Qxb4+ Kg8 4.Nxa8
>>>  +-  (1.99)   Depth: 3   00:00:01
>>>1...Kf8 2.Bg5
>>>  ³  (-0.30)   Depth: 3   00:00:01  1kN
>>>1...Kf8 2.Bg5
>>>  ³  (-0.30)   Depth: 3   00:00:01  1kN
>>>1...Kf8 2.Bg5
>>>  ³  (-0.30)   Depth: 3   00:00:01  1kN
>>>1...0-0 2.h3 cxb4 3.axb4 Bd7 4.hxg4
>>>  ³  (-0.32)   Depth: 3   00:00:01  4kN
>>>1...0-0 2.h3 Bd7 3.hxg4 axb5
>>>  ³  (-0.52)   Depth: 6   00:00:01  8kN
>>>1...0-0 2.h3 Bg3 3.hxg4 Bxg4 4.Nc3 cxb4 5.axb4
>>>  =  (-0.09)   Depth: 9   00:00:01  148kN
>>>1...0-0 2.h3 Bg3 3.hxg4 Bxg4 4.Nc3 Nd7 5.Be3 cxb4
>>>  ²  (0.48)   Depth: 12   00:00:02  2256kN
>>>1...Nc6 2.Bg5 f6 3.dxc6 0-0 4.c7 Qd7 5.Bh4 Bxh4 6.Nxh4 Qc6 7.Nc3 Qxc7
>>>  ²  (0.47)   Depth: 12   00:00:05  3016kN
>>>1...Nd7 2.Nd6+ Kf8 3.Ng5 Bd4 4.Ndxf7 Qf6 5.Rb1 Bg1 6.Nf3 Nb6 7.Qc2 Kxf7 8.Rxg1
>>>cxb4 9.c5 Nd7 10.axb4
>>>  =  (0.06)   Depth: 12   00:00:09  9620kN
>>>1...Nd7 2.Nd6+ Kf8 3.h3 Nb6 4.Qe8+
>>>  =  (0.05)   Depth: 15   00:00:43  86103kN
>>>1...Nd7 2.Nd6+ Kf8 3.h3 Nb6 4.Qe8+ Qxe8 5.Nxe8 Kxe8 6.hxg4 Bd4 7.Nxd4 cxd4 8.Kf3
>>>Ke7 9.c5 Nd7 10.Bg5+ Nf6
>>>  ±  (0.81)   Depth: 18   00:08:19  900744kN
>>>1...0-0 2.h3 Bd4 3.Nbxd4 Nf6 4.Qc2 cxd4 5.Kf2 h6 6.Be2 Nbd7 7.Rf1
>>>  ²  (0.43)   Depth: 18   00:09:40  900744kN
>>>1...0-0 2.h3 Bd4 3.Nbxd4 Nf6 4.Qc2 cxd4 5.Kf2 a5 6.Bd2 Bd7 7.Be2 axb4 8.axb4
>>>  ²  (0.58)   Depth: 20   00:31:35  3409915kN
>>>
>>>(W,  24.09.2001)
>>>
>>>3,409,915,000 nodes.  And not the correct solution.  :(
>>>
>>>(This is Nolot #6.)
>>>
>>>
>>>Slate
>>
>>3,409,915,000<2^32
>>
>>The real test is when the number of nodes is bigger than 2^32
>>
>>Deep Fritz has no problem but I do not know remember other chessbase programs
>>that can  report more than 2^32 nodes
>>
>>Uri
>
>Yes I hope, that in future all engines can handle 64b node counter. Specially
>annoying is Goliath, which can reach 2^32 quite fast because of extreme NPS!
>
>Jouni

DJ7, DF and Crafty all get from 1.5~2.0M nps.  It doesn't take long........

With todays hardware rapidly getting faster, I think it's time to "rethink" how
to display analysis up to at least 1 trillion nodes.


Slate



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.