Author: Robert Hyatt
Date: 07:18:09 08/18/01
Go up one level in this thread
On August 18, 2001 at 09:17:02, Oliver Roese wrote:
>On August 16, 2001 at 10:35:47, Robert Hyatt wrote:
>
>>On August 16, 2001 at 07:55:23, Oliver Roese wrote:
>>
>>>On August 15, 2001 at 00:08:26, Robert Hyatt wrote:
>>>
>>>>On August 13, 2001 at 14:56:32, Oliver Roese wrote:
>>>>
>>>>>On August 12, 2001 at 12:26:09, Bruce Moreland wrote:
>>>>>
>>>>>>On August 11, 2001 at 11:26:31, Oliver Roese wrote:
>>>>>>
>>>>>>>On August 09, 2001 at 12:06:45, José de Jesús García Ruvalcaba wrote:
>>>>>>>
>>>>>>>>On August 09, 2001 at 05:25:37, Graham Laight wrote:
>>>>>>>>
>>>>>>>>>Here's the link:
>>>>>>>>>
>>>>>>>>>http://news.bbc.co.uk/hi/english/sci/tech/newsid_1480000/1480365.stm
>>>>>>>>>
>>>>>>>>>Seems the brain magnetic resonance scanning confirms what we've all suspected -
>>>>>>>>>that GMs tend to use their memory, wheras weak players have to do it by
>>>>>>>>>calculation (the chess computer method).
>>>>>>>>>
>>>>>>>>>The number of patterns a GM is said to be familiar with seems to have
>>>>>>>>>mysteriously risen from 50,000 ("Chess Skill In Man And Machine") to 100,000 -
>>>>>>>>>any idea how that happened, anyone?
>>>>>>>>>
>>>>>>>>>-g
>>>>>>>>
>>>>>>>>It is not clear to me what a "pattern" is, as long as it is not clearly defined
>>>>>>>>the number of patterns can be any number you want.
>>>>>>>>José.
>>>>>>>
>>>>>>>Good point.
>>>>>>>Those patterns probably exists just as dreams of some (bad?) scientists.
>>>>>>>Nevertheless if there is something out there who knows how to identify and
>>>>>>>count these patterns, please tell us about them.
>>>>>>>
>>>>>>>Oliver
>>>>>>
>>>>>>I don't know anything about this cognitive stuff, but here is an example of a
>>>>>>pattern.
>>>>>>
>>>>>>If you are watching a human play chess, and the human (white) has a bishop on
>>>>>>a4, the human will very rarely play b3, even if black has no b-pawn to trap the
>>>>>>bishop.
>>>>>>
>>>>>>This is true all the way up the a-file. If the bishop is on a6, the human isn't
>>>>>>going to play b5.
>>>>>>
>>>>>>When people say "pattern" they are thinking about sexy attack patterns on the
>>>>>>king-side, but there are plenty of little things they strive for or avoid
>>>>>>elsewhere.
>>>>>>
>>>>>>bruce
>>>>>
>>>>>Sure, i know what you are talking about.
>>>>>This is that could be coined as "first-order"-terms (from dr hyatt), in analogy
>>>>>to multidimensional taylorseries.
>>>>>But i personally dont think that these kind of information is stored
>>>>>explicitely, as the name suggests it.
>>>>>Since that would be trillions of patterns to handle.
>>>>>For example i might dont play b3, since i feel/think that
>>>>>this leaves the bishop in bad shape. This is something different
>>>>>than to match a pattern.
>>>>>In a relational database i have no difficulties to count the rows
>>>>>with a simple statement.
>>>>>But the human mind is surely not a relational database.
>>>>>So why do some scientists continue to count pattern??
>>>>>How do they do that?
>>>>>Thats my point.
>>>>>
>>>>>Oliver
>>>>
>>>>
>>>>Humans definitely store chess patterns. De Groot proved it quite nicely with
>>>>a test he ran. If you haven't read the book, and want to hear about the test,
>>>>let me know and I will post an explanation here. But it was definitely
>>>>conclusive that 'patterns' are used in playing chess.
>>>>
>>>Would be nice.
>>>
>>>>one simple idea is the pattern of your area code. It is far easier to
>>>>remember your own 3-digit area code (or your 5 digit zip-code) than it is
>>>>to remember a random 3 digit or 5 digit number. Because it is a familiar
>>>>pattern of digits...
>>>
>>>Familiar patterns are easier to recall.
>>>Ok, but i wonder what does that say in favor of the stored-as-patterns-theory?
>>>
>>>Oliver
>>
>>
>>One of De Groot's experiments dealt with recognizing patterns, in a hidden way.
>>
>>He picked two groups of people. One group knew little about chess. The other
>>group were GrandMasters and IM players.
>>
>>He set up some totally random positions that had nothing to do with real games,
>>ie both kings in check, pawns on impossible squares, etc. He would let one
>>person at a time look at the position for a few seconds, then ask them to go
>>to another board and recreate the exact position. everyone was able to
>>accurately recall the locations of roughly 7 of the pieces, whether the person
>>was a GM or novice.
>>
>>He then set up some normal positions reached during real games. And he repeated
>>the test. Novice still got roughly 7 pieces right. GMs got almost everything
>>right. Occasionally they would get two pieces on different squares, but the
>>"theme" was still there. IE knight pinned on the king by a bishop at either a3
>>or b4, so the GM would mislocate the bishop. It was obvious that the GM was
>>using "patterns" to categorize parts of the board.
>>
>>BTW, 7 +/- 2 is a well-known estimate for short-term memory in a human. We
>>can generally recall 5-9 distinctly different pieces of data. The thing is,
>>what is a piece? For 7 random digits, it might be 7 digits are 7 pieces of
>>data. But if the first 3 are your phone area code, you might recall more of
>>the digits as that 3-digit "chunk" takes a single "block" of STM.
>>
>>Cute stuff...
>
>Thanks for the info.
>i dont say anything against that. It is useful to know, especially for
>programmers. It explains lots of rules for good programming style i know of.
>On the other hand it says absolutely nothing about the question how to _count_
>these patterns.
>
>Oliver
>O
That's why it is _impossible_ to emulate a human's method of playing chess
today. That method is not understood. We know the human recognizes patterns
in a generic way (ie bishop pinning knight on king, no matter where the three
pieces are actually located on the board). But how many patterns are
recognized, and the folded into a single "move choice" is not understood at
all. De Groot simply started the process of understanding how humans play
chess, in a small way. He also tried lots of other tests/experiments as well.
But just because you know that an internal combustion engine works based on the
principle that if you burn something fast enough, you get a big pressure rise,
certainly doesn't tell you all you need to know to build an auto engine. And
the human mind appears to be many orders of magnitude more complex than an auto
engine. Understanding small pieces of the "process" is not anywhere near
understanding the whole thing.
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.