[computer-go] Two-headed problems

[Harri Salakoski]

> Isn't there already sets of problems defined for this sort of thing?
> Probably better to use a standard problem set than going to the
> trouble of making up your own and risking making an error in that
> process (thinking you know what the best move is when you really
> don't).
No slighest idea, hopefully with wiki interface, with rich comments, public 
domain or kind of free usable stuff.
Also xml or sgf formats to download data. Rich sematic "best move", exact 
scores white/black moves first(how many ko threads)
needed to kill group and eye stuff and all rich wariety of connection, eye, 
endgame, joseki and other misc problems.

Otherhands I think it is ok idea to make own test set, even set is very 
small and very limited view for program skills it is best in regression
point of view to check you have not broken anything. But I agree that after 
working with such set enought long time you start wonder maybe somewhere 
exist 10000 problem free high quality set, but I have not found any.

gnugo has it regression set I don't know how good it is as narugo has still 
problems with easiest cases.

t. harri


----- Original Message ----- 
From: "Chris Fant" <chrisfant at gmail.com>
To: "computer-go" <computer-go at computer-go.org>
Sent: Thursday, September 07, 2006 8:35 PM
Subject: Re: [computer-go] Two-headed problems


> Isn't there already sets of problems defined for this sort of thing?
> Probably better to use a standard problem set than going to the
> trouble of making up your own and risking making an error in that
> process (thinking you know what the best move is when you really
> don't).
>
>
> On 9/7/06, Peter Drake <drake at lclark.edu> wrote:
>>
>> Rather than run all of these time-consuming self--play tournaments when
>> selecting parameters for Orego, I'd like to give one or more short 
>> problems
>> and see how often (or how quickly) Orego finds the correct answer.
>>
>> One problem that springs to mind is this:
>>
>> .............
>> ..w.......w..
>> .wBw.....wBw.
>> .wBBw...wBBw.
>> ..w.......w..
>> .............
>> .............
>> .............
>> .............
>> .............
>> ..B..........
>> .............
>> .............
>>
>> With black to play, I think black's best move is to run with the group on
>> the right, aiming for the ladder breaker. Do stronger Go players than I
>> agree?
>>
>> I think this is an interesting problem because it effectively requires
>> deeply searching two "local" problems (the ladders). Another good problem
>> (with the locality more well-defined) would be two separate, strictly
>> enclosed life-and-death problems, where black must kill one of two white
>> groups to win, but only one is actually killable. For example:
>>
>> wwwB.Bwww
>> wwwBBBwww
>> wwwwBwwww
>> w..wBww.w
>> w..wBw...
>> w..wBww..
>> wwwwBwwww
>> wwwBBBwww
>> wwwB.Bwww
>>
>> Thoughts?
>>
>>
>>
>> Peter Drake
>> Assistant Professor of Computer Science
>> Lewis & Clark College
>> http://www.lclark.edu/~drake/
>>
>>
>>
>>
>>
>> _______________________________________________
>> computer-go mailing list
>> computer-go at computer-go.org
>> http://www.computer-go.org/mailman/listinfo/computer-go/
>>
>>
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>