Re: [gnugo-devel] reading patch

[Gunnar Farneback]

Paul wrote:
> after lots of testing i've discovered that the following restrictions
> on which chains try to eat are (probably) the best:
> 
>   - own stone played won't be in atari.
>   - if opponent tries to escape with the second liberty, he won't have
>     more than two liberties (allows to read ladder-like attacks).

I have doubts about the second requirement. This may allow a couple of
pointless moves in the reading tree, pushing the depth limits closer
for the real reading. Not allowing self-atari in these contexts is
almost always a good idea though.

> i used approxlib(), but it might be useful to try is_self_atari() for
> checking the first condition.

Yes, is_self_atari() should be more accurate.

> lazarus:10      FAIL   i'm pretty certain it's not a fail. it's rather
>                        a pass. white T6 saves the entire dragon and
>                        black T8 looks like the only move which might
>                        help (and it played T8). the test might need
>                        revision.

T8 doesn't work at this time. White can force with N16 and S15, then
pull out two stones with O17. After that black has to either let white
make two eyes or lose the entire corner in a semeai. Better to cut off
some white stones and connect all own stones with N16, then aim for T8
in the next move.

The real question with regard to your patch is which tactical reading
results change and whether those are for the better or for the worse.

> trevorb:300     FAIL   previously it passed for a wrong reason (F2 was
>                        valued about 7 points). now it values F1 about
>                        29 and F2 about 27. is F1 really a fail?

F1 looks like it allows a more severe incursion at F3 than F2 does. In
any case this test case is properly solved by
--experimental-connections.

> with the patch the list changed to:
> 
>   Move at K19 attacks L19 with good ko
>   Move at K19 attacks N19 with good ko
>   Move at K19 attacks M18 with good ko
> 
> for me, the second version looks more correct. however, it causes
> great underestimation of K19 move. actually, i think it is bad that
> gnugo never considers K18. maybe a pattern will help?

I agree that the ko results look more correct.

>   - if opponent tries to escape with the second liberty, he won't have
>     more than ONE liberty (atari can't be escaped from),

This is a safer condition.

> there remain these results:
> 
> reading:37      PASS
> reading:170     PASS
> owl:237         PASS
> nngs2:70        PASS

Most notably reading:140 doesn't pass with this variant. It occurs to
me that we should probably use the same type of moves that are
generated by break_chain2_efficient_moves(), and in fact the same
code. This would most likely solve reading:140 but possibly cause
fewer failures in the rest of the test suite.

/Gunnar