Bug in Game Explorer?

rnbqkb1r/pp1ppppp/5n2/2p3B1/3P4/5N2/PPP1PPPP/RN1QKB1R b KQkq

This is the position (fen) after 3. Bg5 from the original post, and in our database, this position has only been reached 11 times, so it's behaving correctly.

You are correct, this is a very simple way to understand it, and I stand by my statement its behaving correctly. The position reached after 2. e4 HAS only been reached 2 times, because it looks like 1. e4 e6 BUT its BLACK to move, which is extremely rare. So, 2 games is correct. However, black is dumb and plays 2. ... e5 and now its white to move with pawns on e4 and e5 which HAS been reached hundreds of thousands of times. So yes, simple example, and yes, its behaving correctly.

D'oh -- I get it.  The results on the screen shown after3. Bg5 are not necessarily for positions reached from Bg5 hence the apparently larger set of stats once you arrive there.  Similar signatures, different results.  The ones shown are correct, irrespective of what position preceded them.

Exactly, thats why the Game Explorer is not the panacea of chess. You still need to use your brain to win. :)  Games can transpose in chess, so even if you click on a move that results in a position with only 10 games in the database, you could find yourself 2 moves later transposing back into a major line where the position has been reached 150k times. The game explorer is semi-useful in finding main line moves, its not going to give you forced wins in all your games, sorry to break the bad news.

Exactly! Move order is ignored, as well as castling rights, so you will find it funny that you can en passant in a move order that normally wouldn't allow it.

the explorer works correctly; you just have to understand what info it's giving you to use it. for every move, it's not telling you how many times that move was played in the database. Instead it tells you how many times it could find the position you would reach with that move in the database (with the proper side to move).

the main advantage to this is that the win and loss percentages are more accurate this way. (for example if you looked at the stats after e3 e6 e4 e5 vs all stats for games that reached that position, the people who played e3 e6 e4 e5 were probably not playing strongly, so their results are not very valuable. when deciding whether you wanted to play 2...e5, it would be more useful to see the overall stats for that position in the database. this applies to all transpositions; more info about results from the possible position are more useful than just statistics about the subset of games that followed your move order).

I owe you an apology jay -- I failed to realize that the behaviour I was seeing was in fact exactly the behaviour I was asking for.  It certainly looks similar to how it would if it were implemented incorrectly (i.e. based on the transition, not the resultant position) and, not recognizing that this would be the case, I assumed that what I was seeing was in fact a product of the incorrect version.

No problem. Glad you understand it now. Now if you can just help to explain it to others and save me some time, cause it is a confusing concept and people get confused by this frequently.

Incidentally, because en-passant is only available on the move immediately following the pawn push they could easily be handled by simply considering fens with different ep rights different positions and simply ignoring the en-passant option when the fen indicates it's not an option.

Castling rights, on the other hand, would be virtually impossible since they survive more than a single move -- I don't see any way to handle them other than to simple ignore them and accept the occasional illegal castling option being presented as is currently being done.

I am apparently too stupid to understand simple concepts, that's fine and I can live with that

This said, I was not expecting not to use my brain when analyzing openings and I was not intending to use Explorer to find a forced win in all my games.

However, I am glad to learn, for the sake of the chess.com population, that the limitation is just in my brain and not in Explorer!

The issue is this:

The position after Bg5 has in fact only been reached 11 times which is what the position you are viewing shows.

Once you click on Bg5, there are 2000+ games shown across a number of positions.  Only 11 instances of these 2000+ games passed through your new current position -- the rest were reached from other similar positions i.e. via transposition, which is precicely what you and I thought wasn't being accounted for.

The game explorer is not looking at only those moves made from the current position, but rather at every move that is legal from the current position and counting up the total number of hits in the game database that match the resultant fen.