[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] testcases
From: |
Evan Berggren Daniel |
Subject: |
Re: [gnugo-devel] testcases |
Date: |
Wed, 26 Feb 2003 21:07:20 -0500 (EST) |
On Wed, 26 Feb 2003 address@hidden wrote:
>
> Evan wrote:
>
> > These are some new testcases from a game gnugo played on nngs. A large
> > collection of mistakes contributed to gnugo losing by 71.5 while taking 6
> > stones from uno 7k*.
> >
> > I've tried to be fairly conservative with my tests while still getting all
> > the major mistakes, but there are certainly more tests possible from this
> > game.
>
> I agree with most of the tests.
>
> Why do the tests use reg_genmove? Apparently this what Gunnar
> intended to used in the regressions when the command was implemented
> in October but nobody (including Gunnar) has been using this command.
>
> If we are going to convert to this command perhaps we should do the
> whole regressions at once.
Ah, I had assumed there was a plan to switch. Doing all at once makes
sense, though. Any reason that doing so would be more complex than
s/gg_genmove/reg_genmove/?
> > loadsgf games/nngs/uno-gnugo-3.3.16-200302230528.sgf 36
> > 320 reg_genmove black
> > #? [M15]
>
> At move 36, there are a couple of things on the board which
> are even more urgent than M15. C6 is very imporant, and W
> is threatening H4, which B might defend in a few different
> ways, for example L4. I took this test out.
>
> At move 128, K8 could also be considered. I added this alternative.
Thanks for the comments.
Evan Daniel