Re: [gnugo-devel] Hoshi patch

[Arend Bayer]

On Mon, 11 Aug 2003 address@hidden wrote:

>
> > I think GNU Go shouldn't play the first jump to start with. See
> > http://mail.gnu.org/archive/html/gnugo-devel/2003-07/msg00279.html
> > for alternatives.
>
> After criticism of the original patch, Evan made a revised patch:
>
> http://mail.gnu.org/archive/html/gnugo-devel/2003-07/msg00252.html

Oops, and we missed adding it to pending patches. Please complain in
such cases to Dan, Gunnar or me!

> I felt this was on the right track after revision. If I recall
> correctly it included some variations of the shoulder hit after
> the one space jump but unlike the original patch did not have
> GNU Go playing the shoulder hit.

Yes.

> It included variations for the second kakari (`B' in the
> message Arend quoted above) but I had some criticisms of
> the variations included:
>
> http://mail.gnu.org/archive/html/gnugo-devel/2003-07/msg00259.html

Which is:
On Mon, 28 Jul 2003 address@hidden wrote:

> One move in this hoshi_keima.sgf I'm not sure is joseki:
> I wouldn't play R17 in the following line because it seems
> to me that this leads to the following position which looks
> slightly better for B.

I think it is commonly seen as joseki. In the ~3500 professional games
I have here, this position arises 11 times, and 9 out of them have
white playing at R17.

> (;GM[1]FF[3]
> RU[Japanese]SZ[19]HA[0]KM[5.5]
> GN[Hoshi joseki variations]
> DT[1999-05-10]
> SY[Cgoban 1.9.5]TM[-];B[pd]C[0];W[qf];C[#kogeima kakari];B[qh]C[j
> ]
> MA[mj];C[#hasami];W[nc]MA[ki];B[pf]MA[li];W[pg]MA[li];B[qg]MA[li];
> W[qc]MA[li];B[pc];W[pe];B[of];W[oe];B[qe];W[qd];B[rf];W[re];B[qf];
> W[pb];B[ne];W[od]
> )

This is exaclty what happened in the majority of the pro games.
Apparently they judged W's territory to be big enough compensation for
B's influence? (I have not looked at the actual games though.)

> Here's the variation that I think is joseki, and a
> refutation of a mistake by B.
>
> (;GM[1]FF[3]
> RU[Japanese]SZ[19]HA[0]KM[5.5]
> GN[Hoshi joseki variations]
> DT[1999-05-10]
> SY[Cgoban 1.9.5]TM[-];B[pd]C[0];W[qf];C[#kogeima
> kakari];B[qh]C[j
> ]
> MA[mj];C[#hasami];W[nc]MA[ki];B[pf]MA[li];W[pg]MA[li];B[qg]MA[li];
> W[pc]
> (;B[qe];W[qc]C[Joseki
> ])
> (;B[qc];W[pe];B[of];W[qd];B[od];W[oe];B[ne];W[qe];B[nd];W[nf];B[og];
> W[rc]C[W has the advantage.
> ])
> )

If you change W S17 to P17, this is what happened in a couple of other
pro games. My guess is that W can only play this if the ladders are
good for W. (And it's still a little complicated after B hanes at N17,
and black has other options.)

Some more variations are
http://senseis.xmp.net/?44PointLowApproachOneSpaceLowPincer.

So what I suggest to do to keep it simple enough:
Take Evan's variation, i.e. let W invade at R17. The above line is
a one-way street. If, on the other hand, the opponent as W
attaches at Q17, GNU Go can just play R15 as you suggest.

Btw, maybe the high approach is even simpler, because then Black can
definitely not hane at R17 after the attachment at Q17.

Arend