Author: Christophe Theron
Date: 21:07:24 02/23/98
Go up one level in this thread
On February 23, 1998 at 16:47:05, Andreas Mader wrote:
>For the first time I must admit that I cannot 100% agree with my Swedish
>friends. If I understood right then Fritz ALWAYS gets big hash tables,
>but the other programs are only getting big tables IF POSSIBLE. This
>means that Fritz has an advantage (maybe only little, but maybe a big
>advantage).
>
>Sorry guys, but this time I do not buy the argument that this advantage
>only brings a few points and/or is not measureable. Its a margin of
>error only in one direction. I think _this_ time it is at least unfair.
>And the SSDF list should not be based on unfair conditions. Period.
In my own programmer's experience, going from 32Mb to 64MB hash tables
will never give you so much elo points. It may even be that you won't be
able to measure the improvement, because it stays well below the error
margin of your rating system.
It's only my experience, and it could be that Frans has found a
wonderful tricky way of taking advantage of big hash tables.
But the most likely in this case is that big hash tables (and it seems
Fritz has got no really bigger HT) have nothing to do with the result.
You can go on arguing about HT size, but I think you are going to loose
your time, while there is something else really interesting to find out
about Fritz5.
For example, the SSDF reports that F5 kills R9 by 14-6. Why? Thorsten
says it knew about that and emailed Ed. Could you find a specific
pattern in the games F5 wins against R9? Can you change the result by
giving Fritz half the size of R9 HT?
Someone else says F5 seems very good in endgames. Could it be the key
point? Did Fritz win mostly in the middle game, or in the endgame???
Many people, including me, believed that more knowledge was more
important when you play long games on fast computers. Has F5 more
knowledge, or is the speed still giving it tactical opportunities? It
could, at first glance, be the point. Look at the results against a fast
program (Nimzo) or a smart tactician (Genius). Fritz doesn't kill them.
Interesting, isn't it?
Christophe
This page took 0.01 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.