[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gnugo-devel] Blunder by current CVS
|
From: |
Gunnar Farneback |
|
Subject: |
Re: [gnugo-devel] Blunder by current CVS |
|
Date: |
Sun, 28 Oct 2001 22:43:20 +0100 |
|
User-agent: |
EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode) |
Teun wrote:
> In the game record at move 43 gnugo played K5.
> In the regression it does not reproduce this and
> plays H5 which is also terrible.
>
> This is so bad, I think a bug may have been introduced
> recently.
Several errors here, but I'm not sure this is something recent. The
most notable error is that the constraint of CD103 is clearly
inadequate, as is the constraint of CD100, which are both matched at
H5. (Whether CD103 matches at H5 or K5 depends on the exact matching
order and is arguably a bug in the pattern.)
Second J9 is much undervalued. This move should have big followup and
reverse_followup values, since it is in fact double sente. Now it only
gets points for saving the stones of the string themselves.
/Gunnar