Author: Robert Hyatt
Date: 12:57:10 04/24/02
Go up one level in this thread
On April 23, 2002 at 17:20:01, Fernando Villegas wrote: >Well, It could be so, but I was answering a post that talked of a long match as >they used to be. In a long match - 24 or so- I guess a GM could have more >chances to tame a monster. BTW, I understand you are the main speakker of the >thesis that even top program does not reach, stll, the level of stronger GM's. >Did you change your mind? >My best >fernando First, clarify what you mean by "stronger GM's"? If you mean at 40/2hr games, then no. Programs may be breaking into the low GM-level now, but only barely... If you mean game/1hr games, this is old news. Almost 20 years ago GMs found my program impossible to beat at game/5... (impossible in the context of a match of 4+ games, not a single game). 5 years ago it was painfully obvious that game/30 time controls were difficult for humans to win. We played a round-robin with 4 GMs and 4 computers at game/30, and the four computers finished in the top 4 places, the four GM players finished in the bottom 4 places. game/60 are also very difficult for humans. They often build up significant advantage at that time control, but then the last 1/3 of the game sees them fall into all sorts of bad blunders due to lack of time. My analysis, based on "hot hardware" might go like this: time control computer rating (FIDE equivalent) game/5 3000+ game/30 2700+ game/60 2600+ 40/2hr+20/1hr 2500+ Take your pick, based on the time control you like. I have _always_ used the 40/2hr time control myself, because from the time I started playing chess in the 1950's, thru roughly 1990 or so, that was _the_ time control for WC matches. That has changed. I watched one FIDE WC event determined by a pair of blitz games, in fact...
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.