The previous discussion (http://www.chess.com/forum/view/premium_members/game-explorer-feature-request) started making a lot of sense.
Games Explorer doesn't work right

we know about this. it's a HUGE project and it's on the list of things to do. it has to do with the FEN string. our next round of opening explorer improvements are in Q1 next year. thanks!

Thanks artfizz - I remembered that topic but couldn't find it. I added a comment there, calling the potential solution "flat FEN" because it collapses several positions into one which are technically different but not _really_ different.
Erik - thanks!

FEN is a nuisance - when I was young and used to play chess *seriously* then old forsythe notation was good enough - and compact. And descriptive was much more descriptive... Well we could read algebraic, though slowly, to follow USSR books which probably used algebraic since October 1917...
Sometimes one feels that gen x and y and z are spoilt brats...
there is a Lennon song about it actually [POST-beatles of course] - called "Serve Yourself" - nowadays it is most readily available on the anthology wonsuponatime. Wait, I'll put the lyrics on my lyrics-blog.
I want to start this up again - Games Explorer doesn't work right.
Here is an example. After e4 Nc6 Nf3, it says there are 1,648 games where black plays ...e5. But if you look at e4 e5 Nf3, it says there are 245,471 games where black plays ...Nc6. What is wrong?
I know what is wrong - the Games Explorer considers the positions different because of two things. First, the "50 move counter" is not the same for both positions. Second, when black plays ...e5 there is the (theoretical) possibility that white can capture en passant (though it doesn't apply in this particular position).
In a more technical sense, it is searching all games that have the same FEN string. I would argue that the only extra information which is important is an en passant capture, along with castling or no castling. Other than those two things, it should be searching "by position only".
The above example is sort of obvious. Many people know that if they want to look up e4 e5 Nf3 Nc6, that is the usual move order. But other examples are not so simple. For example, how many people would know that after the following moves in the Caro-Kann Panov-Botvinnik attack
1.e4 c6 2.d4 d5 3.exd5 cxd5 4.c4 Nf6 5.Nc3 e6 6.Nf3 Bb4 7.Bd3 dxc4 8.Bxc4 o-o
The white move 9:o-o leads to exactly the same position as the following moves from the Nimzo-Indian (after white recaptures with 9:exd4)
1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.e3 d5 5.Nf3 o-o 6.Bd3 c5 7.o-o dxc4 8.Bxc4 cxd4
Yet Games Explorer considers the positions different, because the last capture is on white's eighth move in the Caro-Kann line, but the last capture is on white's ninth move in the Nimzo-Indian. The Caro line gives 718 times that white has played o-o, with a 31.2% winning % for white. The Nimzo line gives 1,107 times that white plays exd4, with a 32.2 winning % for white.