[no subject]

[Heikki Levanto <heikki@xxxxxxxxxxx>]

To: computer-go@xxxxxxxxxxxxxxxxx
Subject: Re: Ranking Pattern Matches
Newsgroups: lsd.compgo
Organization: LSD - Levanto Software Development
X-Newsreader: TIN [version 1.2 PL2]

Mousheng Xu (xu@xxxxxxxxxxxxxxxxx) wrote:

: Hi, Smarties,
: 	Suppose you have an empirical pattern database, each pattern has its
: number of instances found in the collected real world. If two patterns have
: been searched out that match against a board, suppose:
: 	1. The two patterns are of different lengths (e.g., 5 stones vs. 4 stones);
: 	2. The shorter pattern is not a subpattern of the longer one;
: 	3. The shorter pattern has a larger number of instances in the database
: than the longer one.

: 	How do you rank the two patterns? 


That must depend on what you use the patterns for!

Generally, I feel that pattern matching is a shortcut to reading, and if the
match is not perfect (as in your case) you will have to take the long way
and read the situation out.

I would hesitate using the frequency in "real world" as a guide. That
reflects so many aspects, mostly the player's confidence that they can
handle the situation as it is. For example, it is not common to see a
bulky-five shape in actual play, both players recognise the situation
earlier, and never play it out that far. Only if a program has the same
level of recognition, confidence, and skills to match, is it any good to
stop at that point!

- Heikki 


--
Heikki Levanto  LSD - Levanto Software Development   <heikki@xxxxxxxxxxxxxxxxx>