Magnus Persson wrote:
> Quoting Petr Baudis <[EMAIL PROTECTED]>:
>
>> Please note that pachi1 had a rather embarassing bug of starting the
>> random playouts with wrong color (so if the last tree node was black,
>> the playout would start with black as well). pachi2 has this bug fixed;
>> the ELO rating is still not settled, but so far it seems that the impact
>> of this has been about 20-30 ELO in the 110k playouts area, which seems
>> surprisingly little to me.
>
> It just shows that MC evaluation is robust, and this is because it is
> relative between moves and not an absolute evaluation. As long as the
> same bug applies to all moves it should not be that important. If the
> playouts are light i guess it matters even less.
Yes, I agree.   With light playouts I would expect starting with the
wrong move to only be a slight weakening of the program.    

 - Don


>
> -Magnus
> _______________________________________________
> computer-go mailing list
> [email protected]
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to