Author: Andrew Dados
Date: 18:47:42 08/24/99
Go up one level in this thread
On August 24, 1999 at 20:28:19, KarinsDad wrote: >On August 24, 1999 at 19:18:15, Georg v. Zimmermann wrote: > >>Are you talking about Bullet against other progs or against humans? >> >>And by the way : I bet Fritz5 is strongest, cause it is a very fast searcher, >>much faster than Hiarcs and Genius. > >I get confused on this fast searcher is better at speed chess issue. > >From my understanding, Fritz5 (.32) handles about 10 times as many nodes per >second as Hiarcs7.32 (give or take on the same hardware). > >To me, this would mean that Fritz5 would get 1+ ply deeper, but well below 2 >ply. So, Fritz5 would go down about 7+ ply on a fast system in 1 second and >Hiarcs7 would go down about 6 ply in 1 second. Does this really correspond to a >REAL significant ability considering that it's only one more ply? > >Doesn't 1 ply only correspond to about 50-70 elo (when you are talking so few >ply) and couldn't that be recovered in a lot of cases due to the heavier >evaluation in a slow searcher (i.e. there are other considerations than just >material gain)? To me, it would seem like the fast searcher would not have THAT >big of an advantage. > >KarinsDad :) From what I read/see in Fritz,CM, my engine, 10x more nodes means 2+ plys deeper search (roughly speaking BF of 3 in middlegame). Also many sources indicate that 2x more nodes (or 2x faster hardware) make up about 60-70 elo points. However his applies to *same engine* only. For 2 different engines node-count comparison is moot (different extensions, knowledge, qsearch, pruning, whole world). -Andrew-
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.