Ticket #122 (new defect)

Opened 6 years ago

Last modified 3 years ago

Impressive Blunders

Reported by: alain Owned by: gnugo
Priority: normal Milestone: 3.9.x
Component: regressions Version:
Severity: normal Keywords:
Cc: patch: no

Description (last modified by alain) (diff)

Some terrible moves lose a won game.

game 33990 move 48 50
At level 0, 5, 10, 12, 15 GNU Go 3.6 and 3.7.9 are always wrong.
GNU Go 3.4 passes the second test (connect a important stone in atari)

game 10655 move 52
at level 12 OK, gg379 connect
at level 13 it play tenuki and gets cut 

$ gnugo-3.7.9-ref --level 13 -l 10655.sgf -L 52
white (O) move D1
$ gnugo-3.7.9-ref --level 12 -l 10655.sgf -L 52
white (O) move J7

Attachments

33990.sgf Download (814 bytes) - added by alain 6 years ago.
terrible white blunder at move 48 and 50
cgos-blunder.tst Download (199 bytes) - added by alain 6 years ago.
test file
10655.sgf Download (867 bytes) - added by alain 6 years ago.
terrible tenuki at move 52 and level 13

Regression Results

Attachment Rev. PASS FAIL Nodes Status

Change History

Changed 6 years ago by alain

terrible white blunder at move 48 and 50

Changed 6 years ago by alain

test file

Changed 6 years ago by alain

terrible tenuki at move 52 and level 13

comment:1 Changed 6 years ago by alain

  • Description modified (diff)

comment:2 Changed 6 years ago by gunnar

  • Component changed from source to regressions

In game 33990, move 48 looks fine to me. Black can't do anything with the J8 stone after cutting. Move 50 invites a difficult to read semeai. I'm not sure who would win it. Move 52 is definitely bad.

comment:3 Changed 6 years ago by alain

Regressions with 1500 semeai nodes:

ninestones:570 FAIL H1 [A9|B10|D11|F11|J15|A16]

ouch! H1 is very bad, pure blunder, just one more dead stone. How is this possible ?

comment:4 Changed 3 years ago by gunnar

  • Milestone changed from 3.8 to 3.9.x
Note: See TracTickets for help on using tickets.