When I see overemphasis of the center, it looks to me as if the bot has not
understood how easy it is for opponents to break through sketchy diagonal
lines. Tweaking the playouts might fix this tendency. "when you see weak
diagonal lines, try to break them with standard sequences" might be encouraged.
Or, more generally, "try reducing moves."
Strong players seem to adapt their tactics to the stones on the board. They are
like a golfer with a diverse bag of clubs, who knows which one to swing.
Terry McIntyre <[email protected]>
Unix/Linux Systems Administration
Taking time to do it right saves having to do it twice.
>________________________________
> From: Olivier Teytaud <[email protected]>
>To: [email protected]
>Sent: Thursday, December 15, 2011 5:13 AM
>Subject: Re: [Computer-go] Sketchy names
>
>
>
>
>Sometimes go bots overemphasize play in and for the center.
>>What if inside the bot territory on the 19x19-board is counted
>>not in the traditional way, but give different points different value.
>>Example: points in the central 7x7-square get value 0.9 each (or
>>7/8=0.875). All other squares get the normal value (=1).
>>How would an MCTS bot with such a setting play?
>>
>
>I'm pretty sure this has been tested in the early times of MCTS without
>progress
>in terms of ELO. No idea of the behavior in qualitative terms; just it was not
>stronger
>than the initial bot. It was something like a linear combination between
>standard counting,
>and counting emphasizing the borders/corners; with the second term decreasing
>to zero.
>
>(Yet, with MCTS, I never know if a conclusion is never really stable, and if
>we should not try again
>e.g. with more simulations per move...)
>
>Best regards,
>Olivier
>
>_______________________________________________
>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
