Re: [gnugo-devel] auto generated regression tests

[bump]

Gunnar wrote:

> The status field of the dragon_data struct is not very accurate no,
> but that is not what the gtp dragon_status function reports. The
> purpose of the function is to return the engine's best estimate of a
> dragon's status, whatever internal data structures are involved. The
> function looks like this:

Then since we have a field called dragon_status that is probably
not a good name for this gtp function. 

> > Probably these tests should be in terms of either matcher_status or
> > owl tests.
> 
> No, dragon_status is perfectly fine. Owl tests are of course fine as
> well, but only really needed when ko results are involved. See
> ld_owl.tst for lots of examples.

After my patch the gtp function matcher_status can be used to obtain the
matcher status and since this is the status used consistently throughout
the engine that seems to me to be the best way to write these tests.

> | # A19 produces seki
> | loadsgf games/century21-2002/goint-gnugo.sgf 237
> | 90 gg_genmove black
> | #? [A19]
> 
> After black A19 white plays A17 and kills unconditionally. If black
> continues with B19, C17, or A16 there's a dead shape and if A18 or
> tenuki, white plays B19 and wins the semeai.
> 
> As far as I can see, black can do no better in the original position
> than play B19 or A17 and get seki or ko. I'm not sure whether either
> of those two is better than the other, though.

Sorry, A19 is obviously wrong and must be a typo. What I think I intended was
A18. But I was wrong that it is inevitably seki.

B19 and A17 produce kos but not as good. For example

B:B19 W:A18 B:A17 and W gets the first ko capture. 

On the other hand if:

B:A18 W:A17 B:B19 W:A19

this is mannenko or 10,000 year ko. W can make seki at any time
by connecting A18. B has no way to make seki. Either player can
start an unfavorable ko. Whichever player starts the ko has to
make the first ko threat.

See: http://senseis.xmp.net/?ThousandYearKo
http://www.msoworld.com/mindzine/news/orient/go/history/oteai2.html

In order to start a ko, W must first fill the
outside liberties at F16, E12 and E14, then make atari at C17
or A16. But B gets the first ko capture.

On the other hand if B wants to start the ko he has to take
the ko, then fill a liberty at C17 or A16. If he does not
take the ko before filling a liberty, W connects and B
has a dead eye shape. Thus W gets the first ko capture.

Dan