[gnugo-devel] Something odd

[a . crossman]

Hi,

I'm currently running GnuGo 3.3.22 compiled for Windows XP with MinGW 
and GCC 3.2.3. I found this odd:

(;GM[1]FF[4]CA[UTF-8]AP[gGo:0.3.9]ST[1]
SZ[19]HA[9]KM[6.5]
PW[Allan Crossman]PB[gnugo 3.3.22]
AB[dd][dj][dp][jd][jj][jp][pd][pj][pp]
;W[qf];B[qe];W[pf];B[nd];W[qi];B[qj];W[pi];B[oi];W[rj];B[rk]
;W[ri];B[ql];W[qq];B[pq];W[qp];B[qo];W[ro];B[qn];W[pr];B[or]
;W[qr];B[nq];W[fq];B[eq];W[fp];B[dn];W[hp];B[er];W[co];B[do]
;W[cq];B[cp];W[bp];B[bo];W[br];B[ao];W[ap];B[aq];W[bq];B[dr]
;W[ar];B[fr]
)

On move 42, Gnugo seems to miss what I think is a killing move, B2. 
That's not the main problem though - I opened up the SGF file with "-
l game.sgf --quiet -L42 -a -t" as per the Tuning webpage, and it 
gives me a totally different move, valued 3 times as highly. 
Specifically, it now wants to play C6 instead of the move it actually 
played, F2.

So I was wondering if you could replicate this discrepancy on Linux, 
and whether it was a bug...
-- 
Allan Crossman 
address@hidden
http://dogma.pwp.blueyonder.co.uk