[Computer-go] Komi and the value of the first move
dailey.don at gmail.com
Wed Apr 7 12:35:07 PDT 2010
2010/4/7 Andrés Domínguez <andresdju at gmail.com>
> 2010/4/7 Erik van der Werf <erikvanderwerf at gmail.com>:
> > On Wed, Apr 7, 2010 at 3:43 PM, Brian Sheppard <sheppardco at aol.com>
> >>>decreases with board size
> >> Since the game-theoretic value is a small positive integer, I don't
> think it
> >> can decrease with
> >> increasing board size.
> > What I meant is that it tends to start with full-board wins on tiny
> > boards (up to max 28 points as far as I know) and then decreases for
> > larger boards. E.g, 9 on 7x7, 7 on 9x9, and maybe 5 on 19x19.
> I think the big komi is an anomaly with very small boards, where white
> player can't even live.:
> 1x1 : Draw, because no legal move (or losing points allowing suicide)
> 2x2 : Draw, both with super ko rule or simple ko.
> 3x3 : B+8 (japanese rules, chinese B+9)
> 4x4 : Dificult, but I think is a draw with japanese rules
> 5x5 : B+24 (japanese, B+25 chinese)
> 6x6 : Probably white can live and black wins but less than 24 points
> Bigger than 6x6 looks like 8 or 9 is near correct komi for odd boards
> (7x7, 9x9, 13x13, 19x19 etc.), but I suspect it will increase very slowly
> bigger boards (much bigger than 19x19).
> I believe that small boards are not representative, because odd numbers
> are much better for black than even numbers, and because when white
> can hardly live, the bigger the board the bigger the territory black can
> Five points of komi on 19x19 looks extremely low, more in the 7-10 range.
It would not surprise me if the value that I believe it asymptotically
approaches is not that low - for instance it may be something like 7-10 for
19x19 but perhaps it will never drop much below that even on 1019x1019
> Computer-go mailing list
> Computer-go at dvandva.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Computer-go