RE: [Bug-gnubg] Match statistics: Error rate per move/decision

[Albert Silver]

All very pertinent comments IMO. Here's my take on the subject.
Personally, I liked the older system for the very reasons Scott pointed
out: the easy reference to lost equity per decision. On the other hand,
this method does allow one to see more precise scores so that scores
better than 0.010 per move get more precise results. Adding a number
wouldn't work as I think it is hard to read. That is only speaking for
myself.

However, I have had two friends, strong players, and power-users not
able to properly figure out the older method, so so much for popular
consensus. They never looked at it as equity lost, even scratched their
heads when I explained, and mentioned the ease of understanding Snowie's
method of already multiplying it by 1000 as done here now. So I'm sure
they much prefer this newer method. 

It seems, from this, that there are significant groups favoring both, so
why not add this in the options? It may be a bit hidden it is true, but
the possibility will be there, and those having clear preferences can
select the one they like best.

                                                        Albert

> > > the overall error rates (for: checker, cube, luck, overall) in the
> match
> > > stats are given as a loss in equity points.  However, the
> corresponding
> > > error rates per move/decision are given in equity points * 1000.
I'm
> > > finding this a little confusing and it makes reading the stats a
> little
> > > painful and I was wondering if this is a bug or if this has been
done
> > > purposely as I've noticed this behaviour only recently.


> > IIRC, this change was made about a month ago, to report error rates
in
> > millipoints per move, similar to the way Snowie reports them.
> >
> > For me it took a little getting used to, but I think I prefer the
new
> way.
> > It's just more intuitive to think of the error rate per move as 5.5
(for
> > example) than .0055.
> >
> > Since most players will range between 2 and 30, the decimal point
serves
> to
> > distiguish clearly between world class play (2.0) and casual play
> (20.0).
> > With the old presentation, seeing the difference between 0.0020 and
> 0.0200
> > was much more difficult, IMHO.
> 
> I must admit that you have a good point here.  On the other hand, the
> difference between 0.002 and 0.020 can also be seen instantly when one
> glances at GNU's verbal rating i.e "World Class" vs. "Casual Player"
or
> whatever.  Personally, I find 0.002 more meaningful because it gives
me
> a better feel of my error rate i.e I know that I lost 0.002 equity
> points per move which is very little and therefore I know I did well.
> However, if I score a 2.0 then the direct connention to equity points
is
> lost at first sight and I'm always finding myself asking "2.0 what"?
> Millipoints I know but the mapping is kind of lost and I need some
time
> to switch.