The knight would mate on b3 except that b3 is guarded by Pa2. Can I get rid of that pawn? I mulled this over for a moment, and finally realized that the only way to displace that pawn was to spend my knight itself: <15...Nb3+ 16.axb3>.

But this idea has merit because I can replace the spent knight with the other knight <16...Nc5!>, once again threatening mate at b3.

<zanshin: Probably because the move played is slightly better. (Rybka 3 on just <25.bxa5> and <25.Nxg5>)

However, my understanding of the situation is that 15...Nb3+ 16. axb3 Nc5 actually makes a single crusing threat with 17...Nb3# which forces the loss of the White Queen to avoid mate. So in that sense the Queen is trapped.

Yet if White selects to avoid the mate with say 17. Rde1, then 17...Nb3+ is a winning Knight Fork which is a kind of double attack.

In classifying the combinations, I mainly consider which category will help me understand and master the tactic. In this case, the feature of the position that enabled me to solve today's puzzle was the awful Queenside castled position with the resulting shaky position of White's King and Queen. So I put it in my "weakened castled position" collection.

But white's position is so cramped that black can afford to give him a free move and still be able to reload the second knight. Here's the position after 16...Nc5:

click for larger view

White has king, knight, queen, rook and three pawns clustered around his king. And he has the move. Yet he cannot prevent a lethal attack from just a knight and bishop.

The reason for this apparent magic? Black's pieces have mobility and can attack key squares; white's pieces have little mobility and cannot control those same key squares. The white king is stalemated, the rook has only one square to move to, the queen is hemmed in by her own pieces and the poor white knight is suffering from the knight's traditional ability to attack squares adjacent to itself. And the pawns are mostly stuck in place. And for all his forces, white has nothing that can control b3.

I see - good question! ... although at that point, White was lost.

click for larger view

He gets my respect by spotting these terrific knight moves after white castles.

After <9. ...f5> , black has a decent king side pawn push going, zeroing in on e4 and trying to disrupt white's strong central pawn structure. After <10. exf5 gxf5, 11. g4>, the kingside is opened up, and white is pretty much commited to castling queenside at this point.

After the exchange <12.gxf5 Bxf5>, black's light-square bishop is now in a superb position with no pawn resistance.

This question of classifying combinations is, in fact, quite an interesting discussion.

You write:

<In classifying the combinations, I mainly consider which category will help me understand and master the tactic.>

This is of course the main thing, and, hence, as you also say, everybody should classify combinations in whatever way they feel help them do just that.

Having said that, however, I still feel that it does make sense to attempt an objective classification of combinations.

At worst, even if such a project didn't lead to an objective classification, it would lead to a thorough analysis of the subject matter, which cannot but help everybody achieve understanding and, therefore, mastery.

The problem with this particular combination is that there is a sense in which it is ONE threat - namely that is ONE move, Nb3, that is the problem.

But there is also a sense in which it is TWO threats - namely that the move has TWO different points.

So what is it? One threat or two threats?

That depends on how we ought to define a threat.

Is the threat to jump to b3 with the knight? Or is the threat to give checkmate?

In some sense both, but the latter must be the main content of a threat. We normally conceptualize a threat of eg scholar's mate as a threat of mate, not as a threat of moving our queen to f7. Of course it involves a movement of our queen, but it also involves a movement of our hand, but it seems that neither of these two movements are central to what is going on. The mate, on the other hand, IS central.

If the threat is 'to give checkmate' then there is also another threat - the threat of a royal fork.

And - as you say - two threats at the same time, according to the chess dictionary, is a double attack.

Therefore, I would classify this as a double attack, while adding that this is a special class of double attack where both threats are executed at the same square. This might sound odd, but is more common that most people realize. Another example of this tactic is the well known combination where the opponents king and queen are both on the back rank and we threaten to invade it with a rook. If he flees with the queen he is mated, if he makes luft he is forked/skewered/pinned, and, regardless, he is finished off with the same move of our rook to his back rank. This particular combination alone is quite common.

Of course someone might want to classify it as a case of 'bad king's position', and of course that would be correct, just as we might want to classify it as a combination that wins material, as a lot of different combinations do, and depending on what you want to achieve such a classification might be best, but I think such a classification is too broad to really UNDERSTAND what is happening. We need to penetrate into the heart of the tactic. Calling it a double attack with both threats on the same square, in my opinion, does just that.

And regarding your suggestion that since it wins the queen it might be classified as a case of a trapped queen, I must object that in that case virtually any combination that wins the queen should be classified as such.

To me, a trapped piece is a piece that is both threatened and dominated/obstructed, ie if it was a king it would be mate. This is clearly not the case here.