Re: [computer-go] Pattern matching - example play

[Vincent Diepeveen <diep@xxxxxxxxx>]

Well what can i say here.

Port the patterns to a format gnugo can read and start playing games with
it i'd say.

At 03:21 6-12-2004 +0100, Frank de Groot wrote:
>----- Original Message ----- 
From: "Tapani Raiko" <tapani.raiko@xxxxxxxxxxxxxxxxx>
>Subject: Re: [computer-go] Pattern matching - example play
>
>
>> If the played games contain all the information, you could equally say
>> that the empty goban contains all Go knowledge. For example, very few
>> games show ladders. It is the possibilities that the positions open up
>> that have most of the knowlegde.
>
>Are you saying that in half a million games played by 40k to 9p, there is
>not a single ladder played?
>
>Because that was the point.
>To say: "Is not" implies not in the entire population.
>My claim was that my 500,000 games contained everything relevant.
>
>Saying that this is by far not so, and using ladders as an example, implies
>that you are saing that half a million games from 25k upwards contain zero
>ladders.
>
>To stick with the religious arguments: "I find that hard to believe".
>
>Furhtermore (and I have been explaining this a few times already): The
>absence of certain patterns in 500,000 games is knowledge too! Very strong
>knowledge that my pattern system uses to achieve such a high pro-prediction
>rate.
>
>Namely when in 500,000 games x 250 moves = 125 million patterns (of a size
>class) occurs zero patternX, then we can assume that perhaps patternX is not
>such a good patter to move on, hm?
>
>
>_______________________________________________
>computer-go mailing list
>computer-go@xxxxxxxxxxxxxxxxx
>http://www.computer-go.org/mailman/listinfo/computer-go/
>
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/