Re: [Scid-users] Functionality to test

[Jeremy W. <jw...@co...>]

I thought I'd add my $0.02; my perspective on ChessDB is purely
as a teaching tool for kids.

> I've made these comments to Pascal before, but might as well make them
> on the mailing lists too, since he has asked here for comments.
> 
> 1) I can see the point in having a weaker player than crafty. I don't
> know how the various engines compare, but Phalanx looks to be okay for
> that purpose.

I think this is critical; none of my kids have a chance against
crafty, even at it's most eviserated mode (and it's hard to cripple
it enough to make it even vaguely close to their league).

I'd even like to see Phalanx added to the default configuration
so it's not necessary to build it (and can we do that with wcrafty too?).

However, I have to admit that I did not realize that the 'Train' button
would let you play against an engine, and I didn't realize that
Phalanx could scale down to be a weak engine (-e 100 is very nice :-/).

That mode is certainly 80% of what I was looking for; the coach mode
I see as an option on top of that.

> 
> 2) I can see the point in having the rating easily configured with a
> slider, as is Pascal's code. Although the number does not correspond to
> ELO in any obvious way - perhaps the slider should be marked 1 to 100,
> rather than 1200 to 2200 unless there is evidence to suggest those
> numbers are about right.

I agree that the slider should be 0-100, unless we can more concretely
determine the ratings (and if we can, then imho, phalanx should have
an option to play at a given rating, so that supercomputers of the future
don't invalidate any assumptions).

But I think that slider is critical as well; knowing that you have
to go into the engine configuration and add a '-e 100' isn't all
that obvious.

> 
> 3) I'm not convinced there is much point in having Crafty flag a blunder
> that Phalanx has made when playing you, *before* you have even thought
> about the move.
> 
> I never play anyone with someone else looking over my shoulder and
> saying "Hay, he (Phalanx) has just blundered. Now think carefully, find
> his blunder and exploit it". To me, that is what this 'Coach mode' is
> doing (at least in its present implementation).
> 
> That sort of thing to me is just not realistic.

I disagree, at least for kids.  At their level, their games swing
entirely on how many blunders they and their opponent make, so
tools to let them learn to make fewer blunders are all to the good.

Think of the current implementation as a game with an increasing
set of "find White's best move" sort of puzzle.
I'd also like the mode to also flag a blunder by the player, though, and give the player
a chance to figure out why they blundered, and go back and do it right.
(I think it's okay to rig a computer
session so that the kid is essentially guaranteed a win, particularly
if that kid has to study the positions and learn a little bit along
the way.  It keeps it fun for the kid, which I find is critical :-/.)

> 
> 4) I have had some stabilitiy issue with this on my Sun SPARC, but they
> are not always reproduable. Some problems Pascal could see when he used
> my Sun remotely, but the last ones I found, I can't reproduce.
> 
> 
> It's an interesting idea, but overall I'm not convinced of the worth of
> the coach mode as such.

I think if nothing else, if it could be used to push the 'Train' mode out
of the engine dialog and into a more visible place, I think that would be
a good change.  That was one of the first things I looked for in SCID,
and when I didn't find it, I turned instead to xboard.  But it's a pain to
get the game from xboard into scid for analysis later...

My agenda, by the way, is to help get ChessDB to the point where it's
a complete package that I can tell any of my kids (I have 190 kids
in my club, and 45 kids on my 'team') to download and it will be useful
right out of the box.  It's really nice that it's open source and
cross platform; that really works well for my purposes.

Cheers,

Jeremy