Black is threatening a simple rook sac 1 ... Rxc4 2 Nxc4 Bxc4 3 Rb1 d2+ 4 Kg7 Be2 winning, so instead white would play 3 Ke1 Bxb3 4 axb3 Ke5 5 f3 f4! at which point things look difficult for white! So white must try to get out of the skewer, but not lose too much while doing this...

Candidates: e4 (loses to the above), Rxd3...

Whoa! Slow down <gofer> look at the puzzle again!!!! Its black's move - not white's!!! So that means I have it already! :-)

29 ... Rxc4
30 Nxc4 Bxc4 (30 Rxc3 Rc1+ 31 Ke2 Rc3 winning)
31 Ke1 Bxb3 (31 Rb1 d2+ 32 Kg7 Be2 winning)
32 axb3 Ke5
33 f3 f4! (f3 must be played to stop the black king from taking white's kingside pawns)

Now there are three ways for white to defend this...

1) Keep a pawn majority on g and h, but lose control of the centre

34 exf4 gxf4
35 g4 a5
36 h4 b5
37 Kd2 Kd4
38 g5 hxg5
39 hxg5 Ke5
40 g6 Kf6
41 g7 Kxg7
42 Kxd3 Kg6 winning as the white king cannot defend against Pf4 promoting and stopping one of Pa5 and Pb5 promoting

2) Keep a pawn majority in the centre, but have to give it up to get any play as the a and b pawns will force the issue anyway!

34 gxf4 gxf4

35 e4?! Kd4
36 Kd2 a5
37 e5 Kxe5
38 Kxd3 b5
39 Kc3 a5
40 Kb4! Kd4
41 bxa4 bxa4
42 Kxa4 Ke3 winning

3) Keep a pawn majority in the centre, and try to hold on to it, but fail! issue anyway!

34 gxf4 gxf4
35 Kd2 fxe3
36 Kxf3 d2!
37 Kxd2 Kf4
38 Ke2 a5 and its all over...

29...Rxc4

if 30 Nxc4? Bxc4 31 Ke1 (31 Rb1 d2+ 32 Kg1 Be2!) 31..Bxb3 32 axb3 Ke5 33 f3 g4

30 Rb1 ... (30 Ke1 Rc1#)
30.. Rc2
31 Ke1 Rxa2

29...Rxc4 30.Nxc4 - The most critical line to look at first.

30...Bxc4

This threatens not only the rook but 31...d2+ and 32...d1/Q. So on 31.Rb1 d2+ 32.Kg2 Be2 wins. White has no good refutation, so...

31.Ke1 Bxb3 32.axb3 Ke5 and I evaluated the position as winning since black has the infamous outside passed pawn.

That's the most critical line I saw, so I went back and looked at another possibility:

Good question. I think white simply miscalculated and just wanted to liquidate the d3-pawn.

... Bxc4
Nxc4 Rxc4
Rb6+ Rc6
Rxc6+ bxc6

White threatens Rxd3.

The pawn on c4 is hanging but after 29... Bxc4 30.Nxc4 (30.Rc3 b5) Rxc4 31.Rxd3 White seems to have improved his prospects. Therefore, 29... Rxc4:

A) 30.Nxc4 Bxc4

A.1) 31.Ke1 Bxb3 32.axb3 Ke5

A.1.a) 33.f3 g4 34.hxg4 fxg4

A.1.a.i) 35.Kf2 d2 36.Ke2 gxf3+ 37.Kxd2 Ke4 - +, White is in zugzwang.

A.1.a.ii) 35.fxg4 Ke4 36.Kf2 (36.Kd2 Kf3 37.Kxd3 Kxg3 38.e4 Kxg4 and Black's a- and h- pawns win the game) d2 37.Ke2 d1=Q+ 38.Kxd1 Kxe3 followed by Kf3, Kxg3 and Kxg4, winning.

A.1.a.iii) 35.f4+ Ke4 36.Kd2 h5 37.f5 (otherwise 37... h4) Kxf5 38.Kxd3 a5 39.e4+ (39.Kc4 h4 - +) Ke5 40.Ke3 b5 and 41... a4 - +.

A.1.b) 33.Kd2 Ke4 34.Kc3 Kf3 35.Kxd3 Kxf2 36.Kd4 Kxg3 37.Ke5 f4 - +.

A.2) 31.Rb4(c3) d2+ 32.Rxc4 d1=Q+ 33.Kg2 Qd5+ - + [Q].

A.3) 31.Rb1 d2+ 32.Kg2 Be2 33.f3 d1=Q 34.Rxd1 Bxd1 - + [B].

A.4) 31.Kg2 Bxb3 32.axb3 d2 - +.

B) 30.Rxd3 Rc1+ 31.Ke2 Rc2 - +, White is in zugzwang and will lose a whole rook.

C) 30.Ke1 Rc1#.

D) 30.Rxb7 Rc1+ (30... Bxb7 31.Nxc4) 31.Kg2 Bxb7 - + [R].

It seems to win as well.>

Looking over your post, I am curious about some of the candidates you chose: (C) and (E).

For example, after C) 30.Ke1 it's true that 30...Rc1 is mate, but even if that were not the case 30.Ke1 doesn't make an attempt to regain material and is basically non-forcing. Black could even play 30...Rc2 and be ahead. E) 30.Kg2 just gives up material for nothing it seems.

Variation D) seems like a good critical candidate since it attempts to equalize, which I did not consider.

The sub-variations below A.1) didn't seem necessary. But I noticed other kibitzers looked further at 33.f3 g4 so maybe I stopped analyzing too early? I evaluated the position as winning on 32...Ke5 and looked at 33.Kd2 Ke4 but didn't put that on my post. I thought it had to be winning because black's king is where the action is (the white pawns aren't going anywhere), and the outside passed pawn on a7 will easily determine the game.

Maybe you can help straighten out my own analysis.

Indeed, checking it out I found that <30.Nxc4 Bxc4> was safe, because white will have to give the rook back. If white tries to save the rook with 31.Rb1? then 31...d2+! 32.Kg2 Be2 and white will have to sac R for Q and lose.

So, white needs <31.Ke1> to unpin his king and stop ...d2, and then <31...Bxb3 32.axb3 Ke5> (preparing for ...Ke4 to guard the d-pawn) <33.f3> (to prevent that, thus threatening Kd2 & Kxd3) <33...g4!> (white has no time for Kd2) <34.hxg4 fxg4>

Black still has no time for Kd2 (35.Kd2? gxf3 36.Kxd3 Kf5! 37.e4+ Kg4 38.e5 Kxg3 39.e6 f2 40.e7 31.f2=Q+ )

But Black's other tries (35.fxg4 or 35.f4+) appear to leave white in a zugzwang after 35...Ke4 36.Kd2 a5!. White is about out of moves, while black has a couple more tempos to waste with his b-pawn. White will then need to surrender the e-pawn and with it the game.

<29 ... Rxc4
30 Nxc4 Bxc4 (30 Rxc3 Rc1+ 31 Ke2 Rc3 winning)
31 Ke1 Bxb3 (31 Rb1 d2+ 32 Kg7 Be2 winning)
32 axb3 Ke5
33 f3 f4! (f3 must be played to stop the black king from taking white's kingside pawns)

Now there are three ways for white to defend this...

1) Keep a pawn majority on g and h, but lose control of the centre

34 exf4 gxf4
35 g4 a5
36 h4 b5
37 Kd2 Kd4
38 g5 hxg5
39 hxg5 Ke5
40 g6 Kf6
41 g7 Kxg7
42 Kxd3 Kg6 winning as the white king cannot defend against Pf4 promoting and stopping one of Pa5 and Pb5 promoting>

Try 38 h5 in your above continuation; I'll think you will find that white is now winning.

click for larger view

click for larger view

I would evaluate this as a win for Black. Outside passed pawn and all that; eventually, White's king will have to deal with a Black passed pawn on the queenside, allowing Black's king to invade decisively on the kingside. If White tries the f3 blockade again, Black will force an entryway with ...g4 (possibly prepared by ...h5). White might be able to force a passed pawn in the center, but that will not essentially change the course of the game.

If I had been playing Black, I would have been willing to steer for this ending, even without calculating all the resulting variations. The plan is a simple and common one, and I would be confident about dealing with any details as they arose.

It doesn't work as good. 29...Rxc4 30.Nxc4 d2 31.Ke2 Bxc4+ 32.Kxe2 Bxb3 33.bxb3