Author: Uri Blass
Date: 13:12:11 11/10/01
Go up one level in this thread
On November 10, 2001 at 16:02:14, Uri Blass wrote: >On November 10, 2001 at 14:54:04, Thomas Mayer wrote: > >>Hi Christophe, >> >>> So we are both repeating that the comparison was useless? Great. >> >>I subscribe this without any problem... :) >> >>> So I still have to understand what was the point of your initial post... >> >>To show that any engine on Athlon 1333 can not be compared with any engine on >>much lower hardware. Not when you try to get after that some estimate how they >>will do against humans... >>But well, I might be wrong... But it's my opinion... So far nothing show me that >>I am wrong... >>SSDF knows about the problem, I am sure - anyway they have done this 100 ELO >>substraction to have the upper part of the list in some compareable range to >>human ELO... and the problem will occur again... And we both agree that it can't >>be fair to substract again 100 ELOs... >>But what else ? Maybe make seperated lists with only small differences in >>hardware ? I have no idea... >> >>Greets, Thomas >> >>P.S.: The opinion with that a simple alpha-beta engine with qsearch would have >>on Athlon 1200 around 2100 is interesting - this shows even more that comp-comp >>lists should not be compared with human lists... In comp-comp lists with very >>different strength there will be always some engines far below 2100 when we say >>that the best are around 2500 (which is a fair estimation for the best amateur >>engines of the winboard engines, I think) But it is hard to guess if they are >>really below 2100 against humans... > >I have bigger estimate for the best amatuers. > >I believe that the best amatuers are around 2600 on athlon1200 when the >commercial programs are around 2700. > >I assume that most of the times humans play against computer without special >preperation for this discussion. > >I believe that this assumption is correct in most of the cases for humans who >played against computers at the low level. > >Uri I can add that Fritz3 already got an IM norm some years ago when it used p90. In that case I know that the GM's did not special preperations and could get a draw in the best case when some masters did special preperation and could do better. Fritz3 was commercial and I guess that it could do even better without being commercial. I believe that your program has not less knowledge than old Fritz3 and if Fritz3 could be close to 2400 on p90 then I see no reason for you to assume that quark cannot do it. Uri
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.