Re: [gnugo-devel] Proposal for helping in development of gnugo (need advice)

[Gunnar Farnebäck]

Denis wrote:
> First of all, thank you for the work you do for providing gnugo.
> 
> Looking at the task list, I plan to develop a regression test file
> for tesuji. Being a weak player, I would use the book "Tesuji" by
> James Davis.
> 
> I would like to get some advices on this before to invest too much time 
> on this activity.
> 1. Are there problems of copyright that should prevent the use of such a 
> file by the gnugo project ?

There are potential problems. The ideas involved in the examples in
the book are unlikely to be original or even copyrightable in the
first place but the exact layouts of the positions and the selection
of material might be argued about.

I'd suggest that you modify the surrounding positions, preferrably
simplifying them (artificially looking positions are perfectly fine
for this use), and at the very least reorder the tests and skip some.
Even better is to mix it up with material from other sources.

See ld_owl.tst for an example of a file with this kind of origin.

> 2. I have already began to build this file checking the results of the 
> GTP commands 'attack' and 'owl_attack'. Is it the right way ?

In most tesuji problems 'attack' and 'defend' are appropriate.
Sometimes owl_attack, owl_defend, connect, or disconnect make more
sense.

> Moreover, I have a third question. In case a problem is detected, what 
> is the right method to use to solve it ?
> - add some patterns in 'attack.db' (and defend)

Rarely.

> - change some code in reading.c

Usually.

> - add some patterns in 'owl_attack.db' (and defend)

Occasionally.

Other alternatives can be to increase the level setting or a specific
search depth parameter, changing code in read_connect.c, or maybe
adding patterns to atari_atari.db.

/Gunnar