Author: Amir Ban
Date: 05:53:23 11/14/97
Go up one level in this thread
On November 13, 1997 at 11:45:43, Chris Whittington wrote: > >On November 13, 1997 at 11:02:53, Amir Ban wrote: > >>On November 13, 1997 at 10:44:20, Chris Whittington wrote: >> >>> >>>On November 13, 1997 at 10:14:35, Amir Ban wrote: >>> >>>>You need both speed and knowledge, unfortunately for some, but >>>>fortunately for the field. Computer chess is interesting because it's so >>>>damn hard. >>>> >>>>The IBM concept of playing 3-4 silly moves per game but still winning on >>>>the strength of a gadzillion NPS was properly laid to rest in Hong-Kong >>>>and in the 1st DB-GK match. >>>> >>>>On the other hand, playing against someone who outsearches you on every >>>>move is a most unpleasant experience that is not good for your health. >>>> >>>>I thought there were some programs in Paris who played good chess >>>>(Gandalf comes to mind), but just didn't have the horsepower to do >>>>better. >>> >>>This is the old Hyatt chestnut of knowledge-speed trade-off. Knowledge >>>is worth 1 or 2 or 3 or 4 or 5 plies or whatever. >>> >> >>There's no trade-off involved. You simply must be competitive in both. > >The glib response is that GM's aren't competitive on the speed front. > >And anyway, I'm not so sure about this anymore. CSTal was doing 4000 nps >at Paris. Not very competitive in speed, but it performed ..... What >were the nps of the mid-range and the fastest ? Seriously more than >this, no ? > >Chris Whittington GM's are rather good at tactics, sometimes better than computers. Note something important: I'm talking about outsearching = tactical superiority. This is not the same as speed or NPS. NPS is just a ball-park figure anyway, but otherwise it's perfectly possible to waste all your NPS by bad searching methods (ever heard of brute force ?). I don't say you need huge NPS. You need tactical competitiveness, and you can do that by being clever in your search methods, perhaps even with 4000 nps. By rumour, this is what Hiarcs achieves, and to a lesser degree Genius. Remember the DB Prototype-Fritz game where DB played in full time-control a disastrous c4 which Genius4 discards in a few seconds ? My point is that you can get points from someone who consistently outsearches you, but there is no way for you to dominate it. Remember the Rebel-Crafty NPS challenge ? Of course Ed was right, but at no time did he predict that he will win the match. He was realistic. Amir
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.