Author: Robert Hyatt
Date: 20:45:10 09/29/99
Go up one level in this thread
On September 29, 1999 at 23:24:12, Joshua Lee wrote: >Thankyou Professor Hyatt i was trying to ask that but everybody thought i was >saying something else.... I wish Hiarcs looked 14ply:) Well againn is 10 ply >complete 2500elo? if you double clockspeed and get around 60elo then how many >points per ply Thaaaaat's what i was pretty much trying to get answerd . I am >not the clearest speaker, sorry about that. I am glad everyone here is nicer >than the old group i would flip out if some punk tried calling me a nazi just >cause they didn't like how i did things..... anyways thankyou all Conventional "wisdom" has said that one ply is 70-100 Elo points better. That means that for a null-move program, a factor of 2-3X faster hardware gets another ply easily. A non-null-move program either has to be somewhat selective or else tolerates a larger effective branching factor... they may need 4-5-6X faster hardware to get another ply. No idea what Hiarcs' branching factor is, but you can determine this for yourself. do a long search and write down how long to get to 4, then 5, then 6, then N. Subtract 5-4, 6-5, 7-6 and so forth to find out how long each depth took. Divide two adjacent values, smaller into larger, to see how the branching factor looks...
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.