Author: Robert Hyatt
Date: 08:16:13 03/25/03
Go up one level in this thread
On March 25, 2003 at 09:45:15, Chris Carson wrote: >On March 25, 2003 at 09:24:56, Robert Hyatt wrote: >> >>>On March 25, 2003 at 00:45:03, Uri Blass wrote: >>> >>>From the book "how computer play chess" page 192 >>> >>>Table10.1 Correlation between search depth and chess rating. >>> >>>Depth of search 8 year 1980 program Belle Rating 2200 >>>Depth of search 9 year 1986 program Hitech Rating 2400 >>>Depth of search 10 year 1989 program Deep Thought 2600 >>> >> >> >>That's a pretty sensless comparison. IE would you compare 9 plies from >>"the king" to 10 plies from "fritz"? I wouldn't. And the comparison is >>nonsense. Hitech was no faster than Belle. It searched about the same >>speed. Deep Thought was the first (of the three above) that was a quantum >>jump in speed. And with a different search. Also the years (above) are >>wrong. Belle hit 2200 in 1983. Ken was given the "life master" certificate >>from the USCF at the 1983 WCCC event in New York. Hitech didn't hit 2400 >>in 1986 either. And it was a couple of years before Deep Thought hit 2650+ >>as well. >> > >In my opinion, speed and depth are only meaningful when comparing one program >against itself on different boxes (slower vs faster hw) to get an aproximation >of ELO increase for the program on different boxes (but NPS or Depth between >programs is not meaningful). Comparing speed and depth between programs is >useless, whether for comercials or CT/DT/DB. Comparing results (comp vs comp, >Human vs comp) is meaningful. > >Actual ELO and Ranking of GM's that played (ie, ELO of opponents) is the only >method to compare strength of programs (over the board play). Also conditions >of the match/tournament are important. GM's that have months to prepare for a >program (comercial programs) will score better than GM's that can not play the >machine (CT/DT/DB). > >Uri does very good analysis for the qulity of play (also important, shows >program strength and weakness). Except his analysis was for the _wrong_ machine. 1978 belle was _not_ the same as 1980 belle. 1978 speed = 5K nodes per second, 1980 speed was 160K+ nodes per second. Two totally different machines/players.
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.