On 09/03/10 10:41, Fulvio wrote:
Hi!
>> I think the present behaviour is fine in this regard, and I don't think I
>> would opt to use a "filter off" switch very much.
>>
> I believe that the tree window change the filter only for a technical
> reason: when you click on the best games button it search only on the
> filtered games.
Sort off, maybe. In the past you had only one tree and some things that
were perfectly sensible back then might rise some confusion today when
you can have many trees opened.
> What I'm wondering is there a pratical use for some users?
Some loose ideas that come to my mind:
* The "best games" list is restricted in the amount of games displayed
* You might not want to see the best games but those by your opponent
and not those by the best of the best. Of course everyone of us plays
Carlsen, Kasparov and others all the day long, but I usually play much
lower rated players.
* It gives a set of games on the fly that can be further restricted by
AND searches. Maybe just convenient, instead of initiating a position
search first? (Tree can be quite fast if decently cached.)
* Tree sets the filter for base B while I'm working in base A. Switching
to B now just gives what I'd otherwise had to initiate manually first.
> In which case do you find useful that the filter is automatically
> (setting the filter manually requires only 2 clicks) set on the actual
> board position (overwriting your precedent filter)?
I admit that I'm sometimes quite confused by the tree setting the
filter. I could very well imagine that I'd switch that off.
>> Well, that is one point that I certainly think should be changed. My
>> preferences would be to order the games after Year as first priority and
>> after strongest players ratings as second, leaving average rating down on
>> third level in the sorting process.
>>
>> That way the most "modern" approach comes up on top - which is of course
>> also true if some 1400 fellow totally mishandled the position two weeks
>> ago (but then that would be rather obvious).
>>
> for example:
> Game1: White_elo 2700 - Black_elo 2500 - year 2007
> Game2: White_elo 2600 - Black_elo 2550 - year 2010
> Game3: White_elo 2400 - Balck_elo 2400 - year 2010
>
>
> I'm thinking about something like this:
> (white_elo + black_elo) -penalty_year -penalty_elo_diff
As I wrote yestereve, I think that such a calculation might give a lot
better estimate for "best game" than simply the average ELO alone.
However, I wonder if the above formula is already the way to go.
It should be noted that not all databases have ELO ratings included, or
do not use ELO but one of the other rating systems Scid supports. With
your formula, ranking gets difficult or even entirely wrong. Therfore,
one might be better off with adding points and substracting points to a
general rating number, say setting 0 as default and then use a weighted sum.
It introduces some sort of ranking technique here. I like the idea and
it is actually quite extensible. I also think that the formula might get
even a bit longer in the end. Options that come to mind for the ranking
are e.g.
- Annotated games: I think they should get a boost. Usually,
uninteresting games are not annotated. And if a "low rated" game gets
annotations it might definitely be worth a look it could be some
otherwise missed gem. Scid distinguishes between annotations like
Comments (=text), Variations and Annotations (=Nags). All those are
encoded in flags and even their count is present, that is one could
distinguish between silly comments like one commercial database offers
("Annotated by John Doe" instead of using a header field or "From the
circular") and heavily analysed games with n+1 variations.
- Annotators name: taking up an idea by Gerd, I could imagine a list
like "my annotators" (similar to "my player names") and if a game
annotated by one of the annotators listed it gets an additional boost.
(Say "Kasparov" rates higher than "Doe, John". ;)
- Flags: maybe I'm the only one but I use them a lot. This could get
quite fancy here. Say in the opening phase a game flagged as "Opening"
should be more relevant than a game flagged as "Endgame". This even
regardless of it's rating score. This goes in the direction of Vincents
posting provided you have decent (deep-)indexing in your database.
- Source: one might expect to get the worst of metadata from Informant,
but they do quite a selection on what games to include and usually use
high class annotators.
Palle mentiones that the formula get's complex. I wonder if this is
really an issue. Who knows the formulae used by google for it's ranking?
I know the idea what they use, but to the best of my knowledge the
actual formula is top secret. Does this hinder us to use G? So, I do not
think that the complexity of a formlua is an issue in the end. The only
thing that will tell is, if the results improve.
I could also imagine that one can set the weights to some "by us
sensible defaults", but a user can probably change them if he wished to.
This is similar to evaluation criteria for a chess engine. I admit, that
I'm quite happy that I do not have to fiddle them out for each and every
engine myself but that the programmers invested their knowledge and set
some sensible defaults. Still, there might be users out there that tune
the settings to their liking. So why not? Having weights for each
parameter allows to set them to 0, so every scheme could be accomplished
easily. If some user doesn't like a fancy calculation he could set every
weight to 0 except "average rating" and you're back at the algorithm
used now. Sounds sensible to me.
cu
Alexander
------------------------------------------------------------------------------
Centralized Desktop Delivery: Dell and VMware Reference Architecture
Simplifying enterprise desktop deployment and management using
Dell EqualLogic storage and VMware View: A highly scalable, end-to-end
client virtualization framework. Read more!
http://p.sf.net/sfu/dell-eql-dev2dev
_______________________________________________
Scid-users mailing list
Scid-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scid-users
