Author: Bas Hamstra
Date: 01:52:58 02/15/00
Go up one level in this thread
On February 14, 2000 at 12:00:54, blass uri wrote: >On February 14, 2000 at 11:18:21, Bas Hamstra wrote: > >>So far experiments indicate that 2x speed advantage is pretty significant. But >>exactlly how much? Why don't you run a match of 100 games? My bet is the faster >>one wins 75%. > >75% is 200 elo difference and there is no 200 elo difference by being twice >faster. > >If you want 75% you probably need 8x faster hardware. >> >>For 99% certaincy you probably need at least 10x faster hardware, I think. > >I think that maybe no hardware advantage will help to get 99% because of the >fact that there are games when the result is decided by book and not by the >program. > >>Because chess is partly a game of luck and in 100 games the faster engine might >>easiliy get killed once because of a rotten openingline. Unless the speed >>difference is enormous. > >I do not see how speed difference can help if the end of the book line is a >simple win for one side or a simple draw. If it is not a ridiculous book, there are no simple wins. You can get a better position, right out of book. But to *win* that superior position is quite something else. Especially against 2x faster hardware. A friend of mine comes 99% of the time out of opening with serious advantage, using Crafty with his own backsolved and handtuned superbook. But some advantage is not enough, when you play against 4x faster hardware. And the game is long... >I assume that the level of the slow machine is the same and that the level of >the fast machine is the only variable that is changed. >Of course if you do the slow machine slower it can miss a simple tactics or >on time but the interesting question is what happen if the > >If you want to get 99% results you need a good book when there is no simple >or simple win out of book(using a book not from pgn file is a good idea) >but even in this case x10 times advantage is not enough for 99% > >Uri I agree that 99-1 is an extremely hard to get result. Probably you are right and even 10x speed advantage is not enough. Regards, Bas Hamstra.
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.