Re: [computer-go] Pattern Matcher

["Vlad Dumitrescu" <vlad_dumitrescu@xxxxxxxxxxx>]

From: "Frank de Groot" <frank@xxxxxxxxxxxxxxxxx>
> I thought I had explained several times how my pattern matcher works. It
> first collects the 8 million most frequent patterns used in contemporary Go,
> based on counting the frequency of patterns in a population of
> 125,000,000,000. Then it establishes the correlated (relative) whole-board
> values of those patterns by statistical analysis.

Hi,

Please forgive my ignorance and let me ask a honest question myself.

You say that your pattern matched 85% of the plays from an unseen game. I think
it is a remarkable fact.

What concerns me is the rest 15%. From 120 moves, this is about 18 moves. Did
you compare them with what was actually played? Were they better or worse?

I'd also make a comment. I also was into pattern matching (not at the level
discussed here, of course) and I thought it would be a great thing. But later I
realized one thing: a program using this technique is doomed to be a step back
from human players. It has no creativity, it can't invent new things [*]. This
isn't meaning it isn't useful - but that it is only a part of the answer, not
THE answer[**].

Given enough breakthroughs like that, I think we will see major improvements
when someone will assemble them together into a playing machine.

best regards,
Vlad

[*] Let's assume a perfect pattern matcher that had as base data all games ever
played up to 1900 (recorded or not). I'm sure it would be very, very strong.
Would it stand a chance against Go Seigen, Takemiya or Lee Chang Ho? There's a
lot of never-seen-before stuff in their play, and I hope that the next 100 years
will bring up just as many innovations, and the next 100 even more so!

[**] Well, we all know THE answer is 42, but what was the question again? :-)

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/