RE: [Bug-gnubg] Intelligent Analysis

[Albert Silver]

> >             I have been thinking. Right now one of the great boons
about
> > the engines is how strong they are. Especially interesting is that
they
> > evolved their knowledge by themselves and that way were able go
beyond
> > us. The most frustrating part of it all is that they cannot share
this
> > with us in terms we can really understand, so we are forced to
accept
> > they are correct, and then try to explain it to ourselves. Fine, but
it
> > doesn't need to be that way for everything.
> 
> >             Take cube decisions.
> [..]
> > game will be for the match. Your MWC should be X%"
> 
> This kind of information is available from Analyse->Market Window.

Yes, I agree, but the program could highlight this and present it in a
friendly manner for users. I really don't see any major difficulty for
this. Of course, it's as you say: the information IS there if you
consult the right charts, but this way, the program would do this for
the player, and present the relevant information first-hand. Let's
examine the basic cases:

Money game situations

1) My position is simply too weak and even in a money game, doubling
would be a mistake.
2) My position warrants doubling.
3) My position is too strong, and I should play on for the gammon as I
have X% chances to achieve one (that information should be included
BTW).

Match score situations

a) My position is strong enough to double in a money game, but due to
the match score it is not. Of course a) will present the most cases.

        Note that if in a match situation, my position were too weak to
double even in a money game (case 1), it wouldn't tell me it was too
weak for both. It would simply say my position isn't strong enough to
consider doubling yet. Then if it recognized that case 2 was correct
(doubling in a money game) but a) was also true it would present the
information that explained this. It would present that information
verbally (generic script) with the proper statistics. 
        Ex: "Doubling here is a serious mistake. In a Money game it
would be correct, but the match score (one could write a generic text
for the different situations) at 2-away/4-away means that your opponent
will redouble immediately and the game will be played for the match.
Your MWC should be X% before doubling and it is currently only Y%."

b) Doubling is correct.
c) My position is too strong and I should play on for the gammon. One
could work out the cases especially incluenced by match score. There are
a few, but not so many. They change mostly by scores, and not so much by
fundamentally different reasons.

I think that this would be very attractive to users personally. It sure
beats looking up a bunch of charts to figure out why the double was
wrong.

                                       Albert

> 
> > The situations are more than a few, but there are a few that appear
far
> > more often than others, and those initially could be scripted. In
order
> > to not drown the Double analysis window, a button "Explanation"
could be
> > added should the player want it. In time, more cases would be
covered.
> > Perhaps the program can never properly tell me that breaking my
anchor
> > is a catastrophe (though I have my thoughts on this too), but this
is
> > outside the domain of purely evaluation aspects (in other words its
> > weights) and could and should become a part of a teaching tool IMO.
> > Any thoughts?
> 
> It's very hard to implement this. We would have to put logic into
gnubg
> to tell it what positional parameters to look for, since the neural
net
> is a "black-magic" evaluator. 

I understand. I also didn't say this would be too easy, merely that it
would make an attractive feature. First you would need a routine that
idetified the different characteristics, then it would need to find
common points between the moves and their scores.

Ex: the list gives

1) Cubeful 2-ply  13/7* 24/23  
2) Cubeful 2-ply  13/7* 21/20 (-0.018)
3) Cubeful 2-ply  13/7* 6/5   (-0.042)
4) Cubeful 2-ply  24/18 21/20 (-0.135)

Hitting the blot on the 7-pt is clearly obligatory to avoid blundering.
It is also the only common point between the non-blundering moves. The
routine would recognize this common point, the difference between the
scores for not doing it (when there is a common point, otherwise it
keeps quiet) and note that not hitting on the 7-pt is a blunder. After
which 24/23 is the best move with the 1. Again, if it can't explain the
use of the 1, it has said the most important, and this is essentially
filler info. What does this add? Well, right now it just gives a list of
moves and scores. With this it would begin to help identify patterns so
as to help explain what was wrong. We'll still have to figure out the
reason why hitting on 7 is so important, but at least we now know what
to look at as opposed to only "my move was a -0.135 blunder, 13/7* 24/23
was best."

                                        Albert