Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Comments on SSDF by Mr.Diepeveen

Author: Drexel,Michael

Date: 06:54:12 03/05/04

Go up one level in this thread


On March 05, 2004 at 08:37:51, Vincent Diepeveen wrote:

>On March 05, 2004 at 08:23:30, Albert Silver wrote:
>
>>On March 05, 2004 at 03:54:57, Afzal Siddique wrote:
>>
>>>Hello All,
>>>
>>>http://www.aceshardware.com/forum?read=105063596
>>>
>>>Afzal
>>
>>I take it that this accusation can be substantiated?
>>
>>"DIEP is not at SSDF because i refused to pay them. it is a paid list. When you
>>do not ship them a computer or 2 (say $10000) you cannot join in that list in
>>such a way that they directly test your program. Instead they wait then till
>>there is newer books from other programs that can kill you. I refused to do
>>that. I would say now to Karlsson (head of the list) who asked me those 2
>>computers: "You are a surgeon, you can pay for your own hardware"."
>
>When i asked a few years ago whether i could put diep at SSDF,
>Karlsson shipped me email that he could only garantuee diep to be quickly on the
>list if i would ship him a computer or 2. "As we are in big lack of hardware". I
>then heard from someone else he is a surgeon.
>
>>The following comment also needs a bit of clarifiying:
>>
>>"Further there gets used a lot of tricks in SSDF. The protocol to play other
>>programs has a 100 tricks in order to fool you.
>>
>>Example if you play single cpu with an UCI engine against fritz8 at a single
>>cpu, it will eat 80% system time on average versus your engine 20%."

AFAIK the SSDF does not test on a single cpu ?!

Furthermore your statement is complete nonsense.
UCI engines and Chessbase engines get always between 96% and 99% system time in
the recent Fritz GUI in matches with ponder OFF.

Your statement is simply wrong for ponder on as well, but
results with ponder ON on a single cpu are not interesting anyway.


>All these tricks are there. Just try it yourself and measure.
>
>>The way it is phrased, one could easily end up thinking the SSDF actually uses
>>single CPUs to play the matches, as opposed to two separate computers.
>
>I'm sure they play basically single computer now. Either that can have 2
>processors or 1.
>
>>I mean, you'd think that all UCI engines were immediately condemned, instead of
>>the no. 1 in the SSDF list actually being a UCI version of an engine as is the
>>case.
>
>Shredder interface has its own tricks. Sorry features.
>In general less than chessbase but he has little choice of course.
>
>I do not know how SSDF tests shredder, whether they use the native shredder book
>in shredderclassic or shredderbook in fritz + uci engine.
>
>In the latter game it would get toasted as it would get 1 MB hashtables.
>
>If not then chessbase has given order to not trick shredder with the 1MB trick.
>
>I did not test that latter. Perhaps some can try here. It's interesting to know
>the motivation and tricks applied by chessbase. I try to keep updated.
>

Show me evidence that UCI engines get only 1 MB hashtables in engine matches.
The 1 MB hashtable problem definitely exists but I never saw it happen in
engine-engine matches.

Michael

>For now there is just too many tricks. And majority we cannot check even,
>because the games are not there.
>
>>                                           Albert



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.