Petr Baudis: <[email protected]>:
>  Hi!
>
>On Tue, Jul 06, 2010 at 05:50:37PM +0900, Hideki Kato wrote:
>> I and Hiroshi have put FudoGo and Aya into CGOS 13x13 and now four, 
>> including the anchor, GNU Go, programs are running.  Please join us 
>> if interested.
>> http://cgos.boardspace.net/13x13/standings.html
>> 
>> Since strongest MC bots are reaching the pro level on 9x9 board, I 
>> think the next target will be 13x13.  In the Computer Olympiad 
>> Kanazawa, a new tournament, 13x13 Go, will also be created.  In the 
>> human Go community in Japan, there is a TV program for bigginers 
>> supported by Nihon-kiin for a year.  Also, Chang Hsu, ex Meijin in 
>> Japan, wrote and told 13x13 is the best for begginers in his books 
>> and TV interviews and is organizing a personal 13x13 Go tournament in 
>> Taiwan.
>
>  On a related note, I have trialed Pachi on 15x15 board for about a
>month on KGS and I became a huge fan of the size - the advantage is that
>classical strategic concepts and joseki are much more visible on 15x15
>and 13x13, yet the game is much shorter than 19x19. So, I'd like to
>encourage others to also try out this board, perhaps when demoing their
>bots on KGS or such. :-)

I agree 15x15 is also interesting for begginers.  One interesting(?) 
point for both board sizes is that the center is so narrower than 
19x19 that simple MC bots (including my FudoGo :-) which just enclose 
center area cannot be strong.

>  (For program development, I'm not sure if we do need another mid-step;
>with good playing level reached on 9x9, it may be enough to just
>continue tuning directly for 19x19. We have at least the Shodan Go Bet
>to aim at. ;-)

I, now, think mid-step is necessary, as a watcher of Zen19 :).  I'm 
just a 2k player but can understand there are many problems left to 
compete with strong human players.  Recognizing a double ko, for 
example, may require some higher-level analysis.  Currently no good 
way to manage even simple ko fights in MC simulations is established.  
On the hardware side, currently (by the root parallelism, I mean) an 
HPC cluster doesn't help much but improves just a few stones on 
19x19.

Yamato told me that Zen will have no chance to win the Shodan Bet 
even using HPC clusters.

Hideki
-- 
Hideki Kato <mailto:[email protected]>
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to