Re: burte force and knowledge

[Darren Cook <darren@xxxxxxxxxxxxxx>]

>> It's not that we're stingy, "search" happens to be a technical term in
>
>Has this thread degenerated to dictionary flames?

David was pointing out why people were disagreeing with you - when you say
search that is what most of us on this list will think you mean.

>> In contrast, if you are searching for a T in a field of
>> L's, you will have to scan the display. Similarly in go, if you're
>> looking for a shape, say a crosscut, you'll have to actively scan the
>> board until you find one

This is what would most likely happen if presented with a board position
from a game you've not been playing. You'd scan through to see what cuts
work, what groups are alive, what is territory and what isn't. Maybe if you
had to make a quick decision you'd scan for important shape moves and play
the first urgent one you find.

>But this *isn't* what happens in practice.  There are lots of studies of
>where chess masters look at on chess boards and they direct their gaze
>at the few interesting places.

This is what will happen if you've played the previous moves in this game,
because you've already stored what cuts work, life status of groups, etc.
Also for each you've stored what moves might change that status, so if the
last move wasn't one of them, you don't even need to look at that part of
the board again.

Time for an strongly held opinion: doing this well will be one of the
cornerstones of a strong program.

>OK.  The question is only: how does this accurate judgement happen?
>We know it's not sequential processing of each possibility, and it's surely
>not magic.  Parallel evaluation is the only plausible possibility.

I don't know of any proof that it isn't sequential processing. As far as
I'm able to watch myself think about a move, I seem to think about one move
at a time, but I also jump around a lot - I think about move A for a bit,
then move B, then C, then back to A.

Yoshikawa has done some studies on how humans solve life/death problems
under time pressure (3 secs per move I think). One of his conclusions were
that there is some significant difference in the way a strong dan player
stores patterns compared to a strong kyu player. The dan player has much
more meaning and context in the patterns, and isn't really storing
black-white-empty.

I'm probably mixing in my own opinions there. I'll see if I can find any
online versions of his papers.

Darren