[Top][All Lists]

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

Re: [Bug-gnubg] Suggestions

From: Massimiliano Maini
Subject: Re: [Bug-gnubg] Suggestions
Date: Wed, 28 Jan 2009 17:21:31 +0100

address@hidden wrote on 28/01/2009 13:46:29:

> Misja Alma wrote:
> >
> > 2. Batch rollouts. Both for a complete match (with some option for
> > specifying thresholds; maybe not only errors but also everything which
> > is close, or all cube decisions), and also for multiple positions.
> > Actually when 1. is implemented it would be easier to batch rollout
> > positions since they could automatically be saved.
> >
> Thresholds sucks for doing batch rollouts. On the one hand, suppose that
> you have made a double-whopper-cheese-with-sides kind of error according
> to 2ply. No rollout is ever going to make you right. On the other hand,
> suppose you have some super close non-contact moves. Rolling them out is
> probably a waste of your precious cpu cycles. So both large and small
> thresholds sucks. Marking what you want to roll out is the only
> reasonable way as far as I can see.
> Christian.

I tend to agree with Christian, but it is also true that players like it
very much when gnubg 2ply tells them that they have made a -0.060 error and
then after the rollout, it turns out that the player's move was better or
very very close to gnubg 2ply move. Of course they are not interested in
all the cases where 2ply agrees with them but the rollout tells that they
have made a mistake ... skewed appreciation of + and - I guess :)

reply via email to

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