Author: Robert Hyatt
Date: 16:22:48 06/16/98
Go up one level in this thread
On June 16, 1998 at 14:21:12, Bruce Moreland wrote: > >On June 15, 1998 at 14:42:46, Robert Hyatt wrote: > >>On June 15, 1998 at 14:26:59, Bruce Moreland wrote: >> >>> >>>On June 15, 1998 at 13:17:45, Mark Young wrote: >>> >>>>When they use this as an analogy, at best it s a smoke screen, and at worst it >>>>shows how little they do understand about the theory s they use as an analogy. >>>>It just makes them look silly. >>> >>>I don't think it is necessary to get into relativity to make sense out of it, >>>it's just a matter of converting something tangible into something intangible in >>>order to convert it back, with interest, later. >>> >>>bruce >> >>Except I don't personally believe there is any 'conversion' going on. IE >>if you sac a pawn, to get it back "with interest" later... either you didn't >>really sacrifice it in the first place (it was a long-term tactical shot) or >>else your opponent blundered. But there is no way to sac a pawn, and then get >>it back later, if this is not tactically forced... >> >>It's just dropping a pawn... > >This is going to be scattered, but I have to hurry since I have a kid who wants >me to read her some books. > >I don't think your arguments have a practical use. > >Sure, chess is a perfect-knowledge game with an outcome that is knowable, >although it is currently unknown. > >You can take any position in chess and compute all the way to the end and >produce a result. > >I can rarely use this algorithmic approach, in practice I have to use >heuristics, which have a chance of failure. > >Counting material is a heuristic. > >Looking at features that have been called "positional" by chess authors is >another heuristic. > >The material-counting approach works well, but there are cases where it fails, >and anyone's heuristics, be they human or computer, will be better if they >understand the positional cases, including cases where the material is >imbalanced, but the imbalance is not a definate indicator of the result of the >game. > >In short, any player, human or computer, needs to be able to evaluate material >sacrifices, in terms of the risk of a negative outcome as opposed to the chance >of a positive outcome. In any practical time-frame there will always be cases >that need to be resolved using positional terms. > >But yes, all of my talk about heuristics is ultimately unnecessary, since you >can describe an algorithmic approach to chess. The algorithmic approach >eliminates the concept of a sacrifice. I would also argue that it eliminates >the concept of tactics, there are no tactics, there is only +, -, or =. > >You can't implement an algorithmic approach in such a manner that it will handle >the general case in real time. > >And because of this, I tend to ignore the algorithmic approach entirely. It is >not practical. > >On our planet, chess still has a postional (strategic) component. > >bruce this is true, but you overlooked the point of my post. When my first program played its first move in 1968, it could search 2 plies deep. By 1976 I was up to 5 plies if I tried real hard. Today, 12 is typical. The point is, what was solved by positional twaddle in 1968 or 1976 is now solved by pure tactical search. We still have some positional concepts, obviously, as I have several thousand lines of those in my evaluation code. But, over time, some of 'em will be phased out. IE in 1968 I evaluated forks. I haven't done that since 1976. In 1976 I evaluated pins. I haven't done that in years either...
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.