[Bug-gnubg] Props with match scores

[Albert Silver]

> -----Original Message-----
> From: address@hidden [mailto:bug-gnubg-
> address@hidden On Behalf Of Joern Thyssen
> Sent: Friday, January 10, 2003 2:37 PM
> To: W.Stroop
> Cc: address@hidden
> Subject: Re: [Bug-gnubg] autoplay suggestion
> 
> On Fri, Jan 10, 2003 at 06:05:56PM +0100, W.Stroop wrote
> >
> > ----- Original Message -----
> > From: "Joern Thyssen" <address@hidden>
> > To: <address@hidden>
> > Sent: Friday, January 10, 2003 5:30 PM
> > Subject: Re: [Bug-gnubg] autoplay suggestion
> >
> >
> > On Fri, Jan 10, 2003 at 04:14:14PM +0000, Joern Thyssen wrote
> > > On Fri, Jan 10, 2003 at 10:25:57AM -0500, Gary Wong wrote
> > > > On Fri, Jan 10, 2003 at 01:42:18PM +0000, Joern Thyssen wrote:
> > > >
> > > > I think the only feature missing from the current code to be
able to
> > > > implement this suggestion is to customise the "initial"
position,
> i.e.
> > > > when "set auto game" is on, gnubg should return to some
specified
> > > > position when resetting the board (as if repeatedly playing a
> "prop").
> > >
> > > Yes, but this is the same as recording a rollout.
> > >
> > > What was suggested was to play a, say, 7 point match with *each*
game
> > > starting with the same non-standard position. This doesn't make
sense
> to
> > > me :-)
> >
> > Let me just clarify: I can see why you want to start the first game
with
> > a non-standard position, but why do you want to start game two,
where
> > the score is, say, 2-0, which a non-standard position?
> >
> > Jørn
> >
> > To count the score of the non standard posistion  :)
> >
> > It's also practical if you want to create a game file of positions
which
> > arise after e.g
> > 1:  6 1  5 2
> 
> Maybe this is just hairsplitting between two people not having English
> as their native language.
> 
> The following three scenarios exist:
> 
> (a) money game: play a number of games starting from the same
position.
> This is equivalent to a rollout, thus recording such a game will be
> covered by the rollout recording discussed in another thread.
> 
> (b) match play: play the *same* game (i.e., the same score, and same
> cube value) multiple times. Again, this is equivalent to a rollout.

Sort of. It's common to play propositions and attach conditions to them.
For example, let's say I have a proposition but it is clear that the
position favors one side. The debate is how much of an advantage, as
well as the ability to play it. Suppose I say I think I can win the
position 75% of the time, and of course you disagree. I therefore attach
a value to a win correspoding to my expectation, and if you win, you get
your value. A 3-1 prop. Perhaps we can set the cube accordingly as well,
etc. A number of choices are possible. This is interesting also for
training purposes. I may choose to play the same position against the
bot to see if I can do at least as good as the rollouts/analysis say I
should. As to the match conditions, even that is quite interesting.
Suppose we have a position, in which I severely misplayed it, because of
the match score. For example, it was 2-away 4-away so I had to play
unusually aggressively. I didn't play properly according to score.
Having played too passively at times, or simply confusing aggressiveness
with kamikaze play. I'd like to play this out 10 times at exactly those
match scores to train my ability to play according to score. It would
make a wonderful training tool. But each of those 10 games would start
with the exact same conditions: 2-away 4-away, and GNU would play
according to score as well. It would be a bit different than the
classical money play prop.

                                                Albert


> 
> (c) match play: play a match starting from some arbitrary position,
and
> play the *entire* match starting from there. After the second game the
> score is 2-0, thus the second game should start from the starting
> position with the cube in the middle, and the score being 2-0. I can
> never imagine why you want to start game 2 from some other position
than
> the usual starting position.
> 
> I now assume your suggestion is covered by (a) or (b), which will be
> covered by the rollout recording code (once written).
> 
> It's possible to do record one match at the time of (c) unless you
want
> each individual game in the match to start from some non-standard
> opening position.
> 
> Jørn
> Jørn
> 
> 
> 
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg