24...Ng3+ 25.fxg3 Qxg3 26.Nf4 Bxf4 27.Qc2

You wrote: "there is even a third solution!
Ng3+, fxg3 Qxf3.eg.. Qd2 f2".

Your post is full of typos, for example, not using ... to indicate a Black move, the Black Queen can't take on f3, and more. If I have deciphered your post correctly, you are proposing the same line that is being discussed by <mikhs>, <GHOST19>, and <doglikegroove>. <doglikegroove>'s post seems to imply that in this line:

24...Ng3+; 25.fxg3, Qxg3; 26.Nf4!, Bxf4; 27.Qc2

I couldn't get 24. ... Ng3+ or 24. ... Qxh3+ to work. So time to start fantasising.

I would really like to play the draughts move 24. ... Qh2# jumping over the h3 bishop. Or the warp jump knight move of 24. ... Qxf2+ 25. Rxf2 Nxf2#.

We can't bend the rules of space and chess, but perhaps I can combine the two ideas together, with a vulcan mind meld of the Ng3+ idea that I looked at earlier.

24. ... Qxf2+ (removes the defender of g3). 25. Rxf2 Ng3+ 26. Kh2 Nxe4+ and white can only defer Nxf2#.

When I first looked at ...Qg3 I was looking for a mate but also thinking materialistically. I saw very quickly that the rook at f1 can go and then I saw that the knight can windmill around back to f2. My first thought was actually that this forks the king and queen - hence black is winning on material. Of course, I then noticed that it happens to be checkmate. Even better!

It is crystal clear to me now that ...Qxf2 is the cleanest and quickest way to mate. It is the best move in the position. Many of you have said more-or-less the same.

But I still think ...Qg3 is a very good move and (this is purely subjective) a more interesting move (not a better one) in some respects in comparison with ...Qxf2. I didn't really think about this much until I read <zenpharaohs>'s comments: <The game move (24 ... Qg3) still wins though, but white should reply

25 Nf4 Bxf4
26 Nxf3 Qxh3+
27 Kg1 Ng3
28 fxg3

white is knackered, but has evaded the mate.>

I didn't think intially about white defending by giving away both his knights (i.e., moving Nf4 and then Nxf3). And yes, I agree that objectively this is the best white's move sequence. Subjectively, however, I don't think it is the best move. I think the best thing that white can do after ...Qg3 is to take the queen with the pawn (as happened in the actual game).

Why? For two main reasons. Here is reason one (forgive the length - also I have to do this with 2 posts):

If white plays 25. Nf4 then ...Bxf4 is the obvious reply. Again mate is threatened, but white has also just given up material - in fact black is now a pawn up after 25 ...Bxf4. The reason that 25. Nf4 is played, as far as I can tell, is that the black bishop at f4 blocks the f8 rook. For example, if white immediately plays 25. Nxf3 then play could play as follows:

25 ...Qxh3+
26. Kg1 Rxf3
27. Ne7+ Kh8
28. Nxg6+ hxg6
29. Rfb1 (or elsewhere) Qh2+
30. Kf1 Qxf2#

In any case, white plays 25. Nf4 and black replies ...Bxf4. White is now forced to play 26. Nxf3 and again black's obvious reply is Qxh3+ and white will reply 27. Kg1. I don't think anyone, aside from very inexperienced players, would fail to see these moves. (I admit that I did not look at this in advance when thinking about ...Qg3, but I think that I would have seen the replies very quickly if white had indeed started off with 25. Nf4). Also black is now ahead in material and white really has no counterplay that is realistically possible. In other words, the objectively best moves (Nf4 followed by Nxf3) are also very easy to analyze in terms of the ...Bxf4 and ...Qxh3+ replies. Black comes out ahead in material, and it is just a matter of finding a suitable winning method afterward. So that is the first reason why I think (subjectively) that taking the g3 queen with the f2 pawn is actually the better move - and I bet that 90% of players would take that queen, even if 100% of computers would not. The other related reason is that a human player - unlike a computer - would know that there is a very small chance that a black opponent might not see the entire combination. In other words, maybe black will take the f1 rook but perhaps black might not see the ...Nxe4+ move. I admit that there is a very small chance of this happening (and in the game black even sacrificed a piece with 21 ...Bxg4 to eventually get to the puzzle position) but faced with the choice of a) giving up 2 knights for really no compensation or b) taking the sacrificed queen in the slim hope that black may not see all of it, then I would choose b). There is also something to be said for the fact that although white loses, white also participates in a more beautiful game if white takes the g3 queen. Also the game will be shorter - and sometimes, this is a good thing. Usually players want to hang on for as long as possible, but how many want to do this when it is completely hopeless? And I think it is completely hopeless to sacrifice the 2 white knights in order to prolong the game. Still, this is subjective on my part. I can see why someone could still prefer 25. Nf4 and 26. Nxf3. And who knows, maybe at some other point in time, I might argue that in fact, there may be a better chance of black going wrong if white plays those knight moves.

In any case, <zenpharaohs> likes 27 ...Ng3. I actually don't think this is the best move, but I do agree that white is winning with ..Ng3. After 28. fxg3, then I think both ...Be3+ or ...Qxg3+ will win for white, but 27 ...Ng3 is a complicated win in comparison to alternative black moves. (I went to my now ancient version of Fritz and set this up to make sure).

Personally, I like my own recommendation 27 ...Qxg4+ (I didn't use any Fritz help to find that one). But <RandomVisitor> gave the following Rybka line:

(-21.47): 24...Qg3 25.Nf4 Bxf4 26.Nxf3 Qxh3+ 27.Kg1 Nf6 28.Rfb1 Nxg4 29.Qf1 Bh2+ 30.Kh1 Nxf2+ 31.Qxf2

But I sort of satisfied myself that the knight sac 24...Ng3 wins.

It forks the K+R, and it cannot be declined since 25.Kh2 allows ...Nxf1+ followed by Qxf2 and mate coming right up.

So, 25.fxg3 Qxg3 (threat:Qh2#), and white cannot stop ...Qh2# by guarding the 2nd rank with his queen, since ...f2! (supported by my Rf8) blocks it.

Hi, <Hector Maluy>. No, you (and I) did not. Just to be sure, I ran Toga II 1.3.1 with <the last move I entered>:

[ply 15/47 time 00:31 value (to White) -1.91]

9.Nxd4 exd4 10.<Qxd4> Nxe4 11.Qxe4 Re8 12.0-0 Rxe4 13.Nxe4

After 24. ... Ng3+; 25. fxg3, Qxg3; 26. Nf4 indeed "blocks both the rook and the bishop".

I looked at this line again, and in a simple-minded fashion, I thought: "Okay, if 26. ... Bxf4 gives White time for 27. Qc2, what happens if 26. ... Rxf4 instead?"

I now think this is the way to win in the proposed alternate line. Black is merely threatening to reactivate the mate threat on h2 by just getting the Rook out of the way, with 27. ... Rxe4. White has to get Queen or Rook onto his second rank to defend h2, but how? Because Black still has the resource of ... f2, unlike in the line 26. ... Bxf4, if 27. Qd2 or c2, then 27. ... f2! This pretty much forces 28. Rxf2, and White loses his Queen.

1: Josef Kourek - Pavel Chrz, Pardubice 1994

click for larger view

Analysis by Rybka 2.3.2a mp : <22-ply>

1. (-#6): 24...Qxf2 25.Ne7+ Kh8 26.Nxg6+ hxg6 27.Rxf2 Ng3+ 28.Kh2 Nxe4+ 29.Kh1 Nxf2#

2. (-#18): 24...Qg3 25.Nf4 Bxf4 26.Nxf3 Qxh3+ 27.Kg1 Nf6 28.Rfb1 Nxg4 29.Qf1 Bh2+ 30.Kh1 Nxf2+ 31.Qxf2 Bg3+ 32.Qh2 Bxh2 33.Ng5 Qh4 34.Rb2 Bf4+ 35.Kg1 Bxg5 36.Re1 Qg3+ 37.Rg2 Qxe1+ 38.Kh2 Qh4+ 39.Kg1 Be3+ 40.Rf2

3. (-6.17): 24...Rxb3 25.Qd2 Qg3 26.fxg3 Nxg3+ 27.Kh2 Nxf1+ 28.Kh1 Ng3+ 29.Kh2 Ra3 30.Nxf3 Nf1+ 31.Kg1

4. (-4.48): 24...Ng3+ 25.fxg3 Qxg3 26.Nf4 Rxf4 27.Rad1 Rxe4 28.Rd2 Re3 29.Rxf3 Rxd3 30.Rfxd3 Qe1 31.Re2

5. (-3.47): 24...Rb7 25.gxh5 gxh5 26.Qd2 Rg7 27.Qh6 Rg6 28.Ne7+ Qxe7 29.Qxh5 Qg7 30.Nxf3 Rh6 31.Rg1

Hi, <Longbrow>. Careful. After two nights out of the cave, you could get sunburn ;>)

Toga II 1.3.1 says White is lost after the B sac, although the best move is to decline:

[ply 19/56 time 10:23 value (to White) -2.65]

22.f3 Be6 23.Qd2 Qh4 24.Kg1