Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / Expanded commentary pallete--Lost in Tuplespace
- By shogi64 (*) [us] Date 2007-09-02 00:58
I use Fritz, but find the diminutive commentary pallet could be much improved.  I hope you either provide a similar standard pallate, but then maybe add a pushbutton at the bottom to give an expanded pallete window.  As you know in Fritz you currently have to enter the symbols by using corresponding keyboard strokes by printing this out from the help.

Along this line, wouldn't it be wonderful if the eval function would be able to "know" to add commentary symbols into the game, such as "with initiative" or "kingside attack", or "controls the center" etc.  when annotating a game, "auto-annotate". 

In answering another forum member through private mail who wished to practice tactics, would it not be possible to enter a tabia or even a deeper mid-game position in the opening you are interested, then similar to "friend mode" in Fritz have the machine play out a line(s)--perhaps user defined as to number, or user defined by a cutoff value in centipawns-- where it plays these inferior lines, allowing winning a pawn, piece, mating attack etc. I am sure this is possible, as I mentioned to the member to get something like "Anthology of Chess Combinations" and how to manually use the machine to find these tactics...but what are computers good for if we don't let them do our homework?  So this would stop short of being an engine-engine match from that position for X number of games, but would display and save lines you might find from a 1700 rated player, then sort them according to topic (as Fritz does under tactics in the database window).
I suppose you should also allow the user to save this in a seperate user defined file.   Now what happens if you want to expand on this work?  Simply select the file and let the machine loop through and continue adding non-redundant moves/lines up to a user defined number of moves; or even allow him to change the parameters of the initial query, something like that.  I suppose this is similar to Fritz playing games backwards and inserting move/commentary BUT with the addition of looping through the games/position/database collection selected.  I think you see the idea, and apology if this has been posted before.  I got this idea in answering question posted by "adele".  It is a direct answer to the stated comment to "play enough games in this opening to where you are comfortable with the middle game positions that arise" but continues from there...what little programming i have done, i know this is do-able and would be a great thing to add!!  Especially if the eval fxn would comment the lines with symbols as mentioned above.  It may be necessary to develop novel symbols?...computer companies often do this when no "standard" exists, or they are expanding a computer language.  I can think of a few examples.  (e.g. I would like for the tabia of an opening to be displayed with a symbol, or even a deeper sub-variation of this line, i.e. you often see the comment "and this is the current CRITICAL position."  As this may change over time, the question arises as to how to update this value if a strong reply is found etc).   Along this line, it may be possible in the opening book to have a seperate column to show why the move is ? or ?! by entering the corresponding symbol from the pallet in this column.  e.g. ?? would have a symbol for drops a piece--maybe something like a small knight or whatever the piece is--perhaps color coded as to whether it is white or black that loses piece, or mate symbol etc.  IF these symbol annotations can be machine generated, it may be an idea to either allow this "column" to be shown or not--i suppose this falls under the heading of a "customizable" opening book. Perhaps this column could be used or offered as an alternative to "marking up the screen with stars and halos" as previously posted, e.g. just enter the symbol for "knight outpost" in the column, or perhaps some combination would work...or user preference is always a good idea.
I believe this is an entirely different scenario than having the computer enter "deep position analysis with branching" where is plays the best moves, here we are asking it to play inferior moves and show their tactical or positional refutation/improvement.  This would be wonderul, as well all don't "Play like a Grandmaster" or "Play like Rybka 2.3.2a" :-)

There are also lines with long forced variations...is there a symbol for this?  Or perhaps you enter a hitherto unknown position and the machine finds that black must play a forced sequence of moves in order not to arrive at a disadvantage...wouldn't it be nice to mark this?  If subsequent analysis shows this to be incorrect the user can always go back and edit.  Similar to this, white may play an "inferior move" but machine analysis shows that in the vast majority of lines unless black plays correctly he finds himself lost in most variations...so although this line may show up as the 3rd best candidate move, the engine discovers that it is a much more complicated line for black unless he knows his way, and perhaps the analysis discovers some wonderful hidden tactical possibilities so white is willing to give this move a try.  So wonder if there is a way for the machine to indicate this in some way...probably on the analysis board opening book (call it the Tabula Rasa="blank slate", ha :-).  Of course this would be useless in situations where the black reply is obvious, i.e. a simple refutation found easily O.T.B.  so this may be a problem in this implementation, but of course user is going to check Rybka's work anyhow isn't s/he?

Many people have alot of screen real estate, or have a video card that allows for 2 monitors.  So having an Analysis Board which can either be docked or undocked may be the solution to the stars and halos.  Keeping a "clean" interface i would think is critical.  It may be nice to allow by user preference to have a blank "opening book" displayed with this analysis board.  Couldn't the deep analysis or whatever options the user chooses be temporarily added here, and a "port to main book" after review be possible?  Perhaps the answer to the "tricky lines" question is answered here as from this position the machine will generate seperate statistics from this position (see below).  Even if you don't have a second screen, allowing an undocked analysis board should be included, as it can always be set to "stay on top" or toggled.  Would it be possible to "blitz" out huge tree from this position and then have the machine automatically screen and check "prune" the best lines and alternates giving engine more time...this gets past the horizon effect...but i am on vacation for 4 days and want to perform a "multiple step" operation...possible.   

I also find it annoying to be playing through a game with the Opening book displayed and find a ! attached to a move, which has subsequently been shown to not deserve this comment!  I think this is similar to the question someone posted of partitioning the database into recent games (by date) in order to find why this line is no longer played (at least in GM praxis) or a subsequent strong reply by black has changed the evaluation of the line.  I do not know if it would be possible for the machine to automatically "clean" or update the book of these commentary symbols, e.g. see critical variation comment above.

It would also be nice for Rybka to play through an opening line, and loop through a selected opening in the opening book to X number of moves deep, looking for novelties which have not yet been played.  I suppose this could be marked with a "N", and then when the user obtains Megabase2010 and 30 games have been played with this new variation, ha, since the machine now finds games played in the database it can then remove the symbol.  (just a thought).   If the N looks good you can move on to the Analysis board above and let it go to work to validate...   (I am not suggesting an operation on an entire 2GB DB here, just user defineable opening).

As far as opening theory and move order, would it be possible to FIND all solid transpositions to position X which may not appear in the opening book?  Slightly annoying to play a move out of order (for which there is no refutation) and see this is a novelty, only to play the next move and have 12 candidate moves appear!

I really like the idea of marking the tabia.  Another forum member posted he enters all the games in the DB into the book, same here, but how annoying it is to have imported a game when some screwball annotator gives e4!! and it ends up in the opening book as such!  Perhaps by marking the (current) tabia position and thus knowing the number of moves to reach this position, it would be possible to have no such commentary symbols appear until the tabia is reached?  I suppose you would have some function whereby you "update book", understand this would be a long operation.

Last point.  I don't know if the games are being held in a relational database or not (i assume so), but there may be some thought given as to this, as I have discussed with a friend who is DB programmer he thinks the distributable desktop version of SQL server may allow for future developments of options--he mentioned things like mining, cubes etc. but I don't know how relevant it is to chess, he just mentioned they are working hard at giving businesses real options at "getting at" the data in different ways, i.e. "business intelligence".   What was that secret room development takes place in?  room 57?  ha.
It may be nice to add some Report functionality?  Some thought would have to be given here...perhaps you can add some Standard reports (much as accounting software does), then allow adding additional columns or data by clicking the yes/no field--modified or "memorized" reports.   I don't know if adding report functionality would add something useful or not.  It actually may be handy tool, as if you implement the "find novelties" in X opening (above), then how do you find them unless you play through the whole book?   Actually the previous suggestion/post by another forum member on displaying % wins by date is a cube, you have added the date dimension.

Now I am thinking of another scenario.  I enter the position and search DB and come up with 30 games played in this sub-variation.  The "recent" book results for black seem promising so I wish to investigate further.  Now I wish to perform operation having Rybka analyze these games backwards (from only X number or user defineable number of moves) backwards to the tabia position (if this is not somehow marked as suggested above this is useless--unless I know the move number of the tabia and have the opportunity to enter this parameter--I do not care to analyze the entire game as Fritz does backwards as I am searching for ideas earlier on in this variation.  So must loop through the games but to save time only from X moves backwards to the tabia; or if not known then enter Analyze from move 19-30.  Now from these results (temporarily saved where and how--perhaps merged into single game) I can choose and open the analysis board and then perform more operations or the multi-step operation.

Another disconcerting phenomena occurs when players turn in a game score of 1-0 but the database only plays up to move 17!!  So there must be a way of deleting these games...either that or perhaps you can hook up with some reputable enterprise such as chesspublishing.com (unless there is someone better out there i don't know).  As they say, garbage in, garbage out...it seems chessbase just puts these games in so they can say you are getting 2.3 million games...irritating.

Best wishes and sorry the long post.  (LOST in the 10^40 tuplespace!!).  CHEERS :-)-- feedback on my crazy ideas welcome!!  And sorry if I do not know capabilities of other software than :-(  Fritz. 
Up Topic Rybka Support & Discussion / Aquarium / Expanded commentary pallete--Lost in Tuplespace

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill