Katonas,
First off, let me say that I can appreciate the amount of effort that must be involved in managing the Tactics Trainer. Addressing all the submissions in this thread alone must be exhausting. Thanks for your comments.
But I have a few comments of my own. This thread could spawn another forum on the nature of the Tactics Trainer, the definition of tactics, and the use of computers in analysis.
I don't have easy access to an engine, but I was able to get Fritz to look at the position after 1. Ra2. It found 1....Rxg2 after a short time (but not immediately), with the only evaluation as mate in 9. I think the next strongest move was +6 or +4, something like that. So even though ..Rxg2 is a forced mate, 1...Qe7 is still considered a winning move? I think most humans would at least look at the capture before the queen move, and I would not consider ...Qe7 a tactic. It takes black two or three moves (..Qe7, ...Qh4 {I assume that's the idea, although maybe Qg5 first is a finesse}, and ..Rg6) before he is even threating to execute what I would call a tactic with ...Qxh2+.
This is just my opinion, and I certainly don't mean to dictate how the TT should be managed, but it's the Tactics Trainer, not What's The Best Move. I can understand not wanting any ambiguity and avoiding complaints from the users, but Rxg2 is the tactical solution, not Qe7. Before computers, a lot of tactics compilations were taken from actual games (Like the Informant series), and they had to be checked by hand (if the editors didn't just accept the notes from the titled players). I also think there is a point at which the intent of the TT becomess lost to the desire to avoid ambiguity.
For example, in the aforementioned Qe7 line, the rook lifts ..Rg6 and ...Rg5 are equivalent. It wouldn't matter to a human in an actual game which one is played. The idea is to play the queen to the h-file, lift the rook on the g-file, play Qxh2+, and then check on the h-file with the rook next move. But for the TT, one of those rook moves would have to be removed from the solution to avoid ambiguity? I think the TT should be able to accomodate such ambiguity, instead of asking submitters to accomodate the TT as it is.
As for removing some of the alternatives, I suppose I can put a pawn on h5 to eliminate the Qe7 line, but that seems very contrived. Why would Black play his pawn to h5 in that position? Again, it seems like we are serving the needs of the TT, rather than the other way around. The other alternatives are not as forcing as the line in the game, and the Bh3 and Bg2 line is definitely the strongest. It's part of a forced mate in 9. How much weaker do the alternatives need to be? And don't different engines give slightly different evaluations? It seems somewhat arbitrary to pick a number like +2 to separate acceptable lines from nonacceptable ones. And if it were up to me, I would stop the solution at ...Qh3. It's clearly won for Black at that point. I certainly didn't see 11 moves deep during the game. I just looked at Bf1 and Bd3, and that was good enough for both me and my opponent. And no reasonable human would sacrifice his entire army just to postpone mate for another six moves. At least, no human I would enjoy playing.
And one final point (phew!) is that this game was played before anyone even knew what the Internet was. So there is no link to provide.
edvinj
http://www.chess.com/echess/game.html?id=16869364
There were many things I had to try to change such as the second move ambiguity. By moving the rook on f1, Nf6 was no longer an alternative winning second move.
64th puzzle http://www.chess.com/tactics/server.html?id=235977
65th puzzle http://www.chess.com/tactics/server.html?id=322589