Check out Symmetric Chess, our featured variant for March, 2024.


[ Help | Earliest Comments | Latest Comments ]
[ List All Subjects of Discussion | Create New Subject of Discussion ]
[ List Earliest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]

Single Comment

Hannibal Chess. Chess with added Modern Elephants (ferz-alfil compound) on 10x8 board.[All Comments] [Add Comment or Rating]
Greg Strong wrote on Sat, Jan 11, 2020 12:42 AM UTC:

Some thoughts ...

The selection of pieces in this game make sense - adding another minor piece that is about the value of a knight but colorbound.  The game didn't seem to play all that well in my experience but I can't really say why.  I do think the opening array makes a huge difference and it is very difficult to evaluate the quality of one.  Chess works so well because the array is perfect - TONS of different openings that are playable.  Thousands, in fact, at least until recently when hundreds and hundreds of years of study has established many as inferior.  I had a thought about the array for this game that I meant to mention to you but don't think I ever did.  How about putting the elephants in the corners?  Then the center eight files are still the same as Chess, greatly increasing the chance that array is good.  The edge pawns would still be protected by the knights (although one does tend to develop them first, which would leave the pawn unprotected, but then again, the most common first move in Chess is p4 which immediately creates an unprotected pawn.)

Another posibility instead of or in addition to the previous thought - upgrade the modern elephants to war elephants (FADs).  Now your new pieces are about the same power as rooks, but colorbound.  If games are more likely to be popular with more popular pieces it might help.  Plus, on a larger board, it may be more appropriate to add rook-level pieces than knight-level pieces.

Finally, I'd say that popularity on GC is not necessarily a great metric of a game's quality.  If a game is very popular on GC, especially in games that don't involve its creator, that certainly is meaningful.  But the inverse, a game is not popular, doesn't necessarily mean much - especially if its creator is not actively promoting it.  There just aren't that many different players here and there are a ton of games.  The numbers are too small to be statistically significant.  Opulent Chess has had decent popularity including many games I wasn't involved with.  I also think it's a good game and its popularity is deserved.  BUT ... it has been around for 15 years, as have I.  If I disappeared shortly after submitting it, as so many of our active users of old have, would it still be popular?  Doubtful.  And what about Unicorn Great Chess?  David Paulowich did disappear almost completely fairly soon after its release, and it has been fairly popular, but I have been here actively promoting it, and David's games in general.  Otherwise?  Don't know.  Certainly we wouldn't have an ongoing GC tournament including both these games if I wasn't here :)  So don't get discouraged - at least not by one data point that doesn't have much statistical significance.

As an aside, for a long time I've been trying to use ChessV to determine how good an opening array is.  This informed the array for Opulent.  But my technique has greatly improved since then and now I am 100% certain that the Opulent array is inferior.  (I'll be officially changing it soon.)  I'll briefly describe the new approach here since the newest ChessV pre-release that I posted will give others the ability to do this too.

Basically, the technique is as simple as running lots of ChessV against ChessV games (only the beginnings needed) with the Variation of Play set to Small.  Save and analyze them and see how much actual variety you get.  The more variety, the better the array.

Without going into too much detail, the implementation of Small Variation of Play ...  The evaluation of a position is based on lots and lots of parameters.  One of the important ones is Piece-Square-Tables (PSTs.)  The is a bonus or penalty for each type of piece occupying each square (player specific.)  You can see the PSTs in ChessV by right-clicking on a piece and selecting Properties.  The Midgame and Endgame tabs will show the PSTs.  (Endgames might have different values than Midgame, obviously - the King should move out and become active in the endgame and pawn advancement becomes more important.)  These PSTs are calculated at the start of a game for each type of piece in the game based on the weighting of various parameters: is it in the small center?  is it in the large center?  how many small center squares does it attack from here (assuming empty board)?  how many large center squares?  how far forward has it moved?  The variation of play makes tiny random adjustments to the weightings of these parameters for the different piece types.  It also makes tiny random adjustments to the weightings to a bunch of other evaluation parameters related to development and other things.  The overall effect, however, isn't totally random.  I'm not just adding a random value to every position as many chess programs do.  And I'm not at all sure that the default weightings of all these parameters is optimal.  The tiny changes, at least when it is set to Small, is about as like to make the engine stronger as it is to make it weaker.