Re: [gnugo-devel] tiny boards

[bump]

> Initially seven testcases fail. Number 1, 101, and 201 should be
> easily solved by revising fuseki.c. 104 and 202 are basically "invade
> eyespace to make seki" problems (currently most easily solved by
> adding patterns). 207 and 209 are possibly okay, but I haven't checked
> and it might depend on the rules used.

Test tiny:209 is quite interesting.

The answer A2 leads to seki, but B can also try D2.  W can
make a ko at A2 but B will have the first capture. In this
case B kills everything. However W can play C4 in answer to
D2 which leads to 10,000 year ko:

(;GM[1]FF[4]DT[2004-12-07]RU[Japanese]HA[0]SZ[4]
KM[5.5]AP[gnugo.el:2.2.8.b5]
PB[Daniel Bump]
PW[GNU Go 3.7.1]
AB[ba][ab][cb][cc]
AW[bb][bc][bd][cd]
;B[dc];W[ca];B[da];W[ac];B[ca];W[aa])

This becomes seki since neither player can fill the ko, but
if we consider the local game (hypothetically as a summand
in a larger game involving several boards) it's not quite
the same.

Dan