Author: Uri Blass
Date: 03:10:02 11/28/02
Go up one level in this thread
On November 28, 2002 at 06:04:30, Sune Fischer wrote: >On November 28, 2002 at 05:53:53, Gian-Carlo Pascutto wrote: > >>On November 28, 2002 at 05:36:08, Sune Fischer wrote: >> >>>Well, as you know there is relation between speed and time, slower machines >>>would need more time. >> >>How are you going to test this way? Time handicaps don't work with >>pondering. > >What do you mean? >Suppose some user has a PII 500 and another P4 2.2, then we find the speed ratio >and grant the PII 500 user that much more time. It can only work in games without pondering. When the engines ponder the faster hardware may get more ponder time and it is not fair. > >>>If we had enough games, this would smooth out and we would get nice average >>>rating performances. It's not like those lists are done with laser precision >>>anyway (statistical noise being a factor and all). >> >>You'll have to try to minimise the noise if you want to draw >>any conclusion from your list except 'Deep Fritz 7 is better than >>TSCP' > >No. >The thing is, hardware doesn't matter! >What do I mean, well let me explain: > >You can make one list on A1200, then that will give you a "good" rating list for >those programs on that machine, but the list might look totally different on P4 >2.5. Ie. bitboarders might have handicap or something. > >So, a nice mixture of hardware doesn't necessarily make a bad list, it just >"averages" a bit more. In fact such a list might be more representative for the >average user not having a A1200 with 256 MB SDRAM and IDE disks at xxxrpm etc. >It's over specified doesn't compare to other configurations anyway (if you want >to take the strict side of the argument). Taking a mixture of hardware is not a good way to decide. It is better to know the rating of the program on hardware A and hardware B and not to know the rating on a micture that is not clear. 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.