Several tactics are used here.

16.block the queen's coverage of g4

18 divert the rook from the f-file

19 fork the king and queen

21 divert the king from the protection of the rook at g3

We have a lot of material surrounding the white king, and clearly it was a matter of finding the breakthrough.

The most promising attack would come from the two knights. In fact, we can easily see an 'almost mate' of: 16...Ng4+ 17.Kh1 Ng3# -- except of course that white has one defense: 16...Ng4+ is met by 17.Qxg4.

Can we take out this defense? I found no way to deflect the queen, but can we block it?

This question brings 16...f3! into view, since it blocks the queen's access to g4. The queen can't take the pawn (17.Qxf3? Nxf3 ), nor can the g2 pawn (17.gxf3? Qg3+ 18.Kh1 Qxh3#).

The only way to stop the knight mating attack is to take the pawn with the rook, which *still* blocks the queen's defense of g4! So, 17.Rxf3 Ng4+ 18.Kh1.

Here, it took me a moment to see the continuation, as the rook now guards against ...Ng3#. However, in a moment I saw the knight fork, and it was over: 18...Ng3+ 19.Rxg3 Nf2+ forking & winning the queen.

Is there a defence I missed? This looks like even stronger than the game continuation.>

19...Qxg3 meets with 20.Qxg4

Hi, <MrMojo>. You are correct. That is the (otherwise) cryptic meaning of my final phrase: <assuring Black of being left with Q for R plus minor piece>, rather than Q for B+N.

If White does not try to swindle with the pseudo-boomerang combination, Black just walks away with Q for R plus minor piece. In practice, the swindle might be the best try to salvage the game, given that defending against Q with R plus minor piece is likely to be both painful and unproductive.

Not so much. :-)

At least the computers agree that 21.Bg5 was inferior to a number of other 'damage-control' moves like Rg4 or Rxd1 or Ne2.

On the other hand, such moves clearly leave black trying to defend a losing game. Not much fun in that.

If I were white, I would also opt for the counter-computer 21.Bg5!? hoping that black (perhaps mentally celebrating his 'winning' fork) might play the lazy 21...Qh5??, after which white wins with 22.Be2!

And if black does find the one-and-only defensive resource: 21...Bg1+!, then you can laugh, congratulate your opponent on a good win, and go do something fun with the time you saved.

So I tried to make 16. ...Qg3+ work. But no matter how many crazy sacrificial lines I looked at, I could not break through. The problem was that the queen was sitting on a square that one of my knights needs.

Rewind, start again. Next I looked at 16. ...Ng4+. Nope, that does not work either, because of 17. Qxg4.

Third time lucky! Ng4 is now on my wishlist, but I need stop the queen recapture first. The only move that does this is 16. ...f3. So let's investigate that. Clearly, the f3 pawn cannot be captured by queen or g2 pawn, so that just leaves the rook. Now knights can get into the action, as per the game.

When hunting Kings, start by looking at how the squares around the K are controlled. Note the dark B controls g1. The K has only h1 and h2, and the N's are setup perfectly for: Ng4+ and Ng3#. Unfortunately, Ng4+ is met with Qxg4. We can't really deflect the White Q, but we can interfere.

Candidate: f3! - interesting move because it blocks the Q, but also opens up the f4 square for the Black Q. I like this move. It is also forcing.

Variations:
16...f3 17.Qxf3 Nxf3
16...f3 17.Rxf3 Ng4+ 18.Kh1 Ng3+ 19.Rxg3 Qxg3 20.Qg1 Bxg1 16...f3 17.gxf3 Qf4+ 18.Kh1 or 18.Kg2

Oops. Qf4+ Bxf4. Well, that's not good.
16...f3 17.gxf3 Qg3+ 18.Kh1 Qxh3#

Time to check. Oh, btw, I love Weeramantry. Simply because his book "Best Lessons of a Chess Coach" is such a joy to read. Bravo.

Interesting with the Q fork. Certainly was correct. My continuation works leaving the N on g4 and continues (from above) 21.Kxg1 Qh2+ 22.Kf1 Qh1+ 23.Ke2 Qxa1 and Black has Q+N+R vs. B+B+N and White has a desolate position.

If you want a reality check, <jheiner>, your idea

<When hunting Kings, start by looking at how the squares around the K are controlled.>

underlies my desire to <widen the attack, to the light squares>.

Looking at this position with Fritz (and I think a few people pointed it out), 19...Qxg3 20.Qxg4 is crushing.

Good way of looking at it. The h5 N (on light) and the Q are where that attack has to come from, since the e5 N and B will have a tough time getting to light squares. So that way of thinking helps to visualize the attack.

Keep the great comments coming. I enjoy reading yours, and afterwards I always think "Darn, I should've WRITTEN DOWN things like the discovered attack on the h5 B or the h3 P pin, because I was sorta thinking that"...but that's not good enough.

I usually post my solution raw, before reading the kibitzing. I blew the N-fork on this one, even though I thought about it as a wishlist early one, I forgot about it later.

Thanks. I recognize my approach (to everything) is very idiosyncratic, so I appreciate the encouragement.

<<>I usually post my solution raw, before reading the kibitzing.>

I admire the courage in a "raw post". (<dzechiel> is amazing for being able to analyze chess in a stream of consciousness, which clearly takes deep experience.) I confine my posts to the "raw", or else admit defeat, because my variations are too detailed to be worth much if they are wrong.

<<>While i'm enjoying getting the main idea of these puzzles, they are really pass-fail to me. You miss a tactic in a real game, you resign.>

While I hold myself to the pass-fail standard, I think I can claim my failures are better than they were 4 months ago. So can you ;>)