Re: [computer-go] Competition of the GO-algorithms

["Mel Alex" <ihtiand@xxxxxxxxx>]

Gunnar wrote:
> There's also a missing reinforcement move around G7 in 050.sgf.

Yes, I have corrected it too.
Because I think, that the competition Database has not contained the missing reinforcement moves.

But it doesn’t mutter. The good algorithm has to calculate result
as the referee on the tournament.
The reinforcement moves have played into competition Database
to unburden the work of the algorithms.
The question is opened about correct results into the Database only.

> How are seki scored? No points at all as in Japanese scoring or points
> for eyes as is common on go servers?

Yes, it's good question. No points at all as in Japanese scoring.

> There can of course be games where missing reinforcement moves are
> noticed by some programs but missed by others, so this criterion seems
> insufficient to verify correctness.

Ok.
New algorithm for the correctness verifying:
If the game has not be solved by any algorithm, it should be checked.
(26 games from 3003)
One missing reinforcement move shifts result for one point.
So, if one or two best algorithms have calculated result different
from correct result no more then 3 points, the game should be checked. 
(37 games from 3003)

Alex Mel
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/