Hallo Darren
1st game I do not know exactly but the 5 first moves are the same as in
game 5. I changed that in the opening book now.
2nd oppenent had black and made no errors
3rd I played the game with a false Komi (valkyria uses 7.5 as default)
and thought it won with 0.5...
4th Unusual opening and opponent had black. Gerhard Knop is really good
at beating Valkyria
5th This opening sequence is typical of Valkyria (the cut at 6.g6 and I
think it is some kind of Bias where it starts a really hard tactical
position early which might not be good enough to win. This tactical
pattern is like a magnet for Valkyria and I fear it is simply too hard
to read out in combination with a second invasion.
My feeling for is that winning as Black is easy. If Evaluation is below
50% as Black I would consider changing the book even if the game is won.
With white strong players often make it really hard to win. I just won a
game on OGS in the titlematch where the opponent made a mistake on ply
59 as black. There was perhaps other moves that there slow but not
loosing. For a very long time though it looked completely lost for
Valkyria.
Often the score look good searching such a position fresh for 5 minutes
but then it starts dropping when Valkyria over time discovers it has a
biased evaluation of the position. It cannot force a win as white, but
it is good at finding a position where hopes are up thanks to biases and
perhaps complexity of the position. And most of time it does end up
winning because of "forced" mistake.
So in my experience 7.0 is the correct komi. But I cannot proof
anything. I also gained a good feeling for how impossible it is to
"solve by searching". Well played games are very long and search only
touches a small part of what is possible even with 24 hours per move.
Magnus
On 2013-12-06 13:32, Darren Cook wrote:
I know it means very little but Valkyria just reached 9Dan ...
http://www.littlegolem.net/jsp/info/player.jsp?plid=22065
Congratulations! The level of 9x9 ch.*.1.1 play on LG is very high,
because everyone playing there is taking it seriously, knows how to
use
long time allowances, and is therefore very unlikely to make an error
in
the last 2/3rds of each game.
I just took a look at the last 5 lost games [1]. 2 as black, 3 as
white...
did you post-analyze to find where (at what move number) you think
your
mistake was?
And any thoughts on correct komi? (At the point I had to stop my own
LG
experiments, I had a slight preference for 7, but couldn't
dismiss 5 as being correct. Given the lack of bias in your losses, do
you feel the same?) [LG uses 5.5 komi]
Darren
[1]
http://www.littlegolem.net/jsp/game/game.jsp?gid=1466860
http://www.littlegolem.net/jsp/game/game.jsp?gid=1466857
http://www.littlegolem.net/jsp/game/game.jsp?gid=1447908
http://www.littlegolem.net/jsp/game/game.jsp?gid=1442668
http://www.littlegolem.net/jsp/game/game.jsp?gid=1428849
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
