bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Re: Long GNU rollout results


From: Joern Thyssen
Subject: Re: [Bug-gnubg] Re: Long GNU rollout results
Date: Sat, 4 Jan 2003 18:25:17 +0000
User-agent: Mutt/1.4i

On Sat, Jan 04, 2003 at 12:07:06PM -0600, Neil Kaz wrote
> 
> Why ?  Perhaps it is bcuz the developers felt SW 4 was allready too slow,
> like most users do. Or perhaps it was bcuz they couldn't or were unwilling
> to figure out how to make it work perfectly ?

I don't think there is any huge difference between evaluating cube
decisions and chequer play decision according to score or not.

Of course the cubeless-to-cubeful formulae are slightly more complex for
match play than money play (although I've not seen Oasya's formluae),
but the expensive part is still evaluating the neural net.

The same can be applied for rollouts.

The only thing I can think of is that they might have some code
specially written for money game that outperforms the match play code by
a mile.

Jørn




reply via email to

[Prev in Thread] Current Thread [Next in Thread]