EVENT WHITE Saijo Masataka BLACK ManyFaces / GoAhead / Participants of EGC2000 RESULT W+R BOARDSIZE 9 DATE 2000-08-09 PLACE Strausberg KOMI 0.0 RULES Japanese SETUP B G7 C3 COM This is a "3-brain" match between White: Saijo Masataka (8 pro dan) versus Black: 3-brain { ManyFaces / GoAhead / Participants of the EGC 2000 } The rules are as follows: The Participants of the EGC 2000 have to choose between two move candidates provided by the computer programs ManyFaces 11 beta and GoAhead. If both programs agree, they have no choice at all. This idea was developed by Prof. Ingo Althoefer (operating ManyFaces). ``M'' denotes ManyFace's moves ``G'' denotes GoAhead's moves ``B'' means both programs agree Game Name: Saijo (8p) vs. ManyFaces / GoAhead / Participants(B) Write Rank: 8p ENDCOM W 1 F3 MARK G@D7 M@G6 COM MF: g6 GA: d7 ENDCOM B 2 D7 W 3 C5 MARK G@D5 M@B1 COM MF: b1 GA: d5 ENDCOM B 4 D5 W 5 D6 MARK G@C6 M@E6 COM MF: e6 GA: c6 ENDCOM B 6 C6 W 7 E6 MARK G@B5 M@E7 COM MF: e7 GA: b5 ENDCOM B 8 E7 W 9 C7 MARK B@B6 COM both programs: b6 ENDCOM B 10 B6 W 11 F6 MARK G@B5 M@F7 COM MF: f7 GA: b5 ENDCOM B 12 F7 W 13 B5 MARK B@B7 COM both programs: b7 ENDCOM B 14 B7 W 15 D4 MARK G@D3 M@H5 COM MF: h5 GA: d3 ENDCOM B 16 H5 W 17 G4 MARK G@D3 M@H3 COM MF: h3 GA: d3 ENDCOM B 18 D3 W 19 C4 MARK G@E4 M@B3 COM MF: b3 GA: e4 ENDCOM B 20 B3 W 21 E2 MARK G@E4 M@B4 COM MF: b4 GA: e4 ENDCOM B 22 B4 W 23 A5 MARK G@E4 M@B1 COM MF: b1 GA: e4 ENDCOM B 24 B1 W 25 D2 MARK G@E4 M@A2 COM MF: a2 GA: e4 ENDCOM B 26 E4 W 27 E5 MARK \t@B7 \t@B6 \t@C6 \t@D4 \t@C4 \t@C5 \t@B5 \t@A5 a@A4 B@E3 COM both programs: e3 From my point of view, this was the decisive point of the game. GoAhead analyzed that the black LL corner may live by playing at a4 with the main variation "Ba4 We3 Ba2" (see variation a). The move at e3 was also assigned a rather large value, because GoAhead wanted to keep the big W group "weak" :-( Therefore, a4 and e3 went into further inspection in the following ... Unfortunately, there was a bug in the life&death reading code! After putting (testwise) the Ba4 stone on the board, GoAhead redefined, by error, the life&death problem around the LL corner including the marked white and black stones. He could not read out this big problem. Consequently, he came to the wrong conclusion that Ba4 does not safe the stones under consideration. After detecting and removing this bug in the definition of live&death problems, the new GoAhead program now plays Ba4 correctly, after which the game may continue. ENDCOM B 28 E3 VAR B 1 A4 W 2 E3 VAR W 1 C2 B 2 A6 W 3 D5 B 4 F5 ENDVAR B 3 A2 W 4 B8 ENDVAR W 29 G5 MARK B@H6 COM both programs: h6 ENDCOM B 30 H6 W 31 H4 MARK B@J4 COM both programs: j4 Black resigns by majority voting. taken from ManyFaces: 11 out of 16 moves = 68% taken from GoAhead: 10 out of 16 moves = 63% ENDCOM