Author: Robert Hyatt
Date: 06:59:45 06/15/98
Go up one level in this thread
On June 15, 1998 at 01:39:33, Hristo wrote: >On June 14, 1998 at 23:11:37, Robert Hyatt wrote: > >>On June 14, 1998 at 18:11:56, Hristo wrote: >> >>>Bruce is correct ... I have my bananas and you have your oranges ... >>>:))) >>> >>>> >>>>No, I don't believe there is a viable alternative to "search" since we >>>>"humans" rely on exactly the same idea. And if *we* have to search to >>>>find tactical wins, the computers are obviously going to follow that >>>>same path. >>> >>>Bob Hyatt = Lord Kelvin ... :))) >>>"The Earth can not be more than 100 000 000 years old. All evidence and >>>physical(thermodinamics) laws confirm this!" >>> >>>There are many places in life where common sence doesn't present the >>>nature of the world. If we stay within the "common sence" we are bound >>>to be misleaded. >>> >> >> >>again, this is nonsense. I am not promoting the flat-earth society, or >>any such thing. In fact, I promote *exactly* the opposite.. I don't >>believe *anything* (chess-wise) that I can't search to a point that I >>understand it. All of this fabrication from you and Thorsten simply >>clouds the issue. Don't put words (or bizarre concepts) into my mouth, >>because they aren't coming from me... >> >>the only point I am discussing here is that "positional chess" is simply >>chess rules that have found to be correct in the past, but are only >>"estimates" rather than "absolutes". I can search to mate. and prove >>it beyond a shadow of a doubt. You can evaluate potential mating nets, >>but with no search, you are going to be wrong some of the time. As >>hardware gets faster, I find deeper and deeper mates with no mistakes at >>all. You rely on positional concepts, and you continue to make >>mistakes. >> >>ditto for other positional concepts like backward pawns, weak pawns, >>open files, 7th rank, absolute 7th rank, bad bishops, and so forth. >>These >>are all "crutches" because we can't think far enough into the future of >>the >>game to see if they are really going to be bad or good in this >>particular >>circumstance... so we rely on witchcraft and superstition and say "it >>was >>bad last time, or the last N times, it is *probably* bad here." And >>that >>is wrong at times. And search can help to improve the accuracy. >> >>To prove the point is easy. You find a program, *any* program, and >>restrict >>it to a 1 ply search, and let's play a match with any other program you >>want, >>say gnuchessX, and we'll even disable it's main evaluation. Can your >>one >>ply search with any evaluation you want avoid a truly dumb search's >>tactics? >>I doubt it. But even if it does, what about smart+1ply vs Crafty+normal >>search? Utterly hopeless for any program I can think of. And I mean >>*hopeless* as in probably a 99-1 result. So search is accomplishing >>something important, and everyone is relying on it heavily... >> > > >Bob you will have your test! I will have my test! If it is not working >you will be happy! If it is working everybody will win(!) and I will be >happy! > >You are missing the point. Please read my prior message in this thread >which appears at the bottom of this one. The evaluation function I'm >talking about has no "prejudice" knowledge about any chess position. >Ditto your examples about the 7th rank, doubled pawns, and weak squares >doesn't apply! > >Lord Kelvin didn't talk about "Flat Earth". Just like you don't. He was >sure, just like you are, that he knows better than the rest because he >has a proof. >Your proof is: "My program is not perfect, but because it is better than >yours I know better. Perhaps I even know the "truth"." >You had your chance to be a participant. Now you'll be nothing more, but >a spectator. > > >Hristo > > I don't know just "because I know". I know because I have *tried*. For a long time. And tactics that can be evaluated via search are accurate, and tactics that must be evaluated by heuristics are error-prone. However, the main point has been that any positional concept you can define can also be defined in terms of tactics and search. And doing so eliminates all the special cases that can fail... >> >> >> >>>Bob, I hope you are not so rigid! It seems as if you have >>>mental-rigomordus in the prime of life(you are very bright, but...). >>>Your argument doesn't contain much value, other than the fact the we are >>>all subjective, and that we evaluate the world through our own >>>experiances. >>> >>> >>>Crafty must evaluate a position based on something else but the search >>>through the tree of possible moves. For instance after 1. f3 crafty >>>plays 1 ... e5. Now how is this happening?! There is no forced win of >>>any sort yet Crafty plays, probably, the best move. It is also highly >>>unlikely that crafty actualy searches all possible moves. Crafty doesn't >>>evaluate the position as 0.00 ... >>>there is some sort of artificial evaluation there. >>>Mister Thorsten point was that instead of searching a huge tree of >>>improbable moves, and then applying some "bogus" evaluation function, >>>one can spend more time "filtering" out the so called "stupid" moves. >>>This "filtering" will depend on another "bogus" evaluation function ... >>>is it possible to have an "objective" evaluation function? >>>Is this within our grasp? Perhaps! >>> >>>Take a distant look at your "tree-of-moves". It looks as if you have >>>created a possible data stream in the time-domain out of a single >>>position. Then you are trying to reconstruct the original position and >>>determine what is the most probable frequency content of this position. >>>The best move keeps or expands your energy!(The problem is you don't >>>look at it this way, although this is what is happening.) The original >>>position already contains the "frequency" infromation, one have to learn >>>how to extract it whithout going to time-domain and back. >>> >>>Best regards. >>>Hristo
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.