Author: Roy Eassa
Date: 13:32:25 10/11/02
Go up one level in this thread
On October 11, 2002 at 16:27:55, Mike S. wrote: >On October 11, 2002 at 15:34:54, Roy Eassa wrote: > >>(...) >>But my thinking is not a simple checkbox or small change to the program, but >>rather a completely different set of algorithms. Basically, a completely >>different program with entirely different logic. > >I guess, an effective search algorithm which is good for the "standard" engine, >will also be good for an anti-human engine. Maybe search algos have to be >different when the evaluation is much bigger, I don't know. > >The programs which are known to use more knowledge, like Hiarcs or Shredder, >reach similar depths with lesser nodes, so they are competive among comps. But I >think if there would be *much* more knowledge included, the loss of depth would >harm not only the performance against comps but also against GMs. GMs are >extremely good in every category of chess, including tactics. > But GMs would lose some games even to 200 MHz computers. My point is that as MHz goes up and up and up, why keep applying it to the search at all? Yes, to beat other programs, but surely not to beat GMs? When a computer beats a GM it is because the GM made a tactical miscalculation, and surely 500 MHz is enough to punish that. So if you have 2500 MHz, why waste the other 2000 MHz on the same thing that you've already accomplished with 500? Instead, maybe devote those extra 2000 MHz to something very different. It MAY be more knowledge in the evaluation, or it may not. Perhaps it's something completely different and not related to the search at all. >It's different with a system like Brutus, where you can - as it was reported - >add much knowledge to the evalutation chip without loosing speed. > >Also there's no general anti-human strategy independent from the humans strength >level I think. Against weakers players below ~2200 for example, programs can >afford a tactical oriented game basically, don't need to be positional giants, >and can afford somewhat more risk than against comps. Some progs play like that >anyway, even without human/comp switch. Main things to care for in addition are >to avoid closed pawn centers and trojan sacs. > >It's probably different against a class of players of 2200 to ~2600, and again >above 2600. I'm not really capable to come to a solid based opinion about that >calibres of players, so to speak. Probably you'd need to be more near their >strength to have an idea. But what's known is (as Bob Hyatt also has explained >often) that those players can spot, and utilize every type of weakness in an >opponents play. I guess they are used to it, from preparations against their >human opponents. > >Kramnik demonstrates that, too. > >From the impressions I have of various programs, I (still) believe that Nimzo is >particularly unpleasant for human players of lesser strength, just like it is >now (i.e. N8) without additional anti-human things. That it's not in the very >top among progs anymore, doesn't matter much in that respect. It's not >speculative IMO, but has an extremely sharp style. It's pieces are like the >tooth of a shark. It seems that it's not good in the endgame compared to others >unfortunately, but that doesn't matter much. You won't see any endgame anyway >when playing against Nimzo :o). > >But OTOH if it should be that Nimzo's tactics are no threat for a GM, it may be >one of the easiest opponents for them. > >But you can't expect to beat a GM with knowledge, when they know more exceptions >than an average player knows rules... :o) > >Regards, >M.Scheidl
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.