I want to point out that the way chessbase does opening classification is worth looking into
In the chessbase prog u can edit or add the positions which classify (and I feel few folks know this) and this prg does it in an eye-blink
once it where 31000+ positions i recall
I know it was posted earlier, but i wanted to bring it up again. But if we can make somthing similar to the idea it would be wonderful.
I once found out that the same position I analysed of my game could be classed as KID or Benoni
so you would find 2 or 3 classification codes for the same position, and it seems like the move order was the decider.
Pretty weird.
I assume that coding-wise it may be nontrivial to implement the following:
* When analyzing a game, update the opening classification at each ply (for the first N plies, starting at move 1).
but that may partially address the problem.
I want to bring this topic up again because it has been bothering me a long time now and should be easy to fix. Every game starting out 1.d4 e6 for example will be classified as a Horwitz defense regardless from it transposing into a French, QGA, QGD or some Indian defense. The reason is that Lichess classifies the opening by the string of opening moves the game starts with I guess. Chessbase and Arena on the other hand classify by position. So all Lichess would need to do is give itself a file with all named opening positions in it (I have a txt file of that nature as it comes with arena) and have the starting X positions checked if they are in that file. If not, give the last found position as the opening of that game.
Nevermind, I somehow misread this thread, I apologize :(
I apologize for bringing this up again, but after thiking and rethinking this topic and after the addition of the new analysis feature it seems more than ever necessary to implement some mechanism into Lichess that correctly recognizes transpositions, since the new feature loses most of its use in relation to opening related statistics when every game starting out 1.d4 e6 is seen as a Horwitz or some General Queen Pawn Opening. It was mentioned before that
a) Problems arise from analysis on various computers
b) Problems arise from games leaving book lines, re-entering them later.
I would like to propose the following solution and would like somebody to comment on that:
Seperate game analysis from opening classification, that is, have the opening classified on a single machine (possibly the local one since it requires minimal resources). Check the first position versus a opening key. If it is in the key, check position two etc. If position x is not in the k, give position x-1 as the opening of that game and stop checking any move thereafter. Thereby it can be guaranteed that book lines will lead to book positions. This could even be implemented in some ways into multi machine analysis as far as I understand. Please do not dismiss my proposal right away and let me know what you think/ why it would/ would not work.
in the chessbase program you can find all move orders and positions leading to the A_E-0-99 classification. look in opening keys, and first install a/ the big key as it is not standard. then look by edit key...
i guess good pract is to let it run several times and / or 10-15 moves deep to handle transitions
We're not talking about chessbase though #18, we're talking about lichess analysis.
I really think #17 has a brilliant idea with having the opening keys checked locally. Recently with the engine upgrade on lichess some of this has been remedied. I see a lot more opening names in my analysis than I ever did in the past on prior accounts. But you are correct, it still handles transpositions poorly. Having it crunch keys locally while the analysis is running across various servers seems like a logical solution, especially since I imagine most people just sit there and wait while an analysis is running anyway.
the topic is about openings and opening classification and transitions.
analysis is diff from classification database vs brute force calc
the better sources to find all the positions is chessbase soft
not diff at all, only tedious work
