21..Qb1?? dropping a full piece (correct was ..Nxe2 22 Rxe2 Bxa1 23 Qxa1); and secondly...

23..Nxa4? Best seems either 23..gxf5 – altho this looks risky, I can’t find anything wrong with it (24 Rd3 or 24 Bd3 are both easily answered by ..Ne4) – or maybe 23..Rfd8, perhaps to anticipate White’s possible threat of f6. The move actually played should’ve dropped a piece for a pawn (24 Rxa4 Rxc2 Kg1), but White either didn’t see this or chose not to go into it (with Rook and three pawns for White’s two Bishops, to me Black still looks better here); and thirdly...

25..e5? (playable seems ..Nxa2, even tho it sidelines the Knight for a few moves. I toyed around with many replies and the best I could find for White was something like 26 fxg6 hxg6 27 Bc4 e5 28 Rg4 Nb4 29 Rxg6+ Kf7 30 Rg4 and now at least three replies – ..f5, ..d5, ..Nxc2 – all seem winning for Black; now back to 25..e5?) After 26 fxe6 fxe6, now White **clearly** should’ve played 27 Rxd6 reaching this position:

click for larger view

Every extra pawn White can regain makes a difference, and the open nature of the position gives White’s Bishop-pair some chances. Altho Black still stands better, she will have to demonstrate some degree of accuracy to force further simplification, for ex.:

<1]> If 27.. Ne2 (with the idea 28 Bxe2? Rxc2 regaining one of the Bishops) 28 Be5 and after what surely seems best, 28..Nf4, now there are two lines:

click for larger view

<1A]> 29 Rd7 Nxd3 30 Rg7+ Kh8 31 cxd3 Rc5 32 Rf7+ Rxe5 33 Rxf8+ and altho Black should win this pawn-up Rook ending, White will make her work hard to do it.

<1B]> 29 Be4?! (now threatening 30 Rd7) ..Rc4! (keeping up the threats) 30 Bxb7 (there is nothing better) ..Rxc2 31 Rd7 Rf5! and Black must ride out only a few more complications, after which the White Bishops’ “play” (counterplay) will have ended.

<2]> Also playable is 27..Ne4 (with the idea 28 Bxe4? Rf1+) 28 Rd7 (28 Rxe6? Nf2+ 29 Kh2 Nxd3 30 cxd3 Rce8 wins) when I believe Black’s best is ..Nf2+! 29 Kh2 (if 29 Kg1? Black *can* safely insert 29..Nxd3 forcing 30 Rxd3 Rxc2 – because White *can’t* play 30 Rg7+?? Kg8 and Black stays a Rook up, for if 31 Rxg6+?? Black has ..e5. Now back to 29 Kh2)

click for larger view

...when now 29..Rf7 is winning. But note that here, after 29 Kh2, Black *can’t* insert 29..Nxd3? because after 30 Rg7+ Kh8 31 cxd3, regardless of what Black plays, White either forces a perpetual or re-establishes material equality, or retains enough play for her slight material deficit:

<2A]> 31..Rc1 32 Bd4! and now either 32..Re1 33 Rf7+ and 34 Rg7+, or 32..Rf5 33 Rc7+ e5 34 Bxe5+ and 35 Rxc1.

<2B]> 31..Rfe8 32 Be5 (I believe 32 Bf6 reaches equality with almost the same following moves) ..h5 33 Rxb7+ Kg8 34 Rxa7 Ra8 35 Rg7+ Kf8 36 Rxg6 Rxa2 37 Rg5.

(but again these two subvariations are IRRELEVANT, because simply 29..Rf7 is winning)

<3]> More clear-cut than lines <1]> and <2]> is 27..Nd5 28 Rxe6 Nf4 29 Re7 Rf7 (obviously forced) 30 Rxf7 Kxf7 31 Be4 b6

click for larger view

...when White’s position again goes rapidly downhill. Black threatens 31..Rc4 (which will win the c2-pawn) and White’s only logical attempt to hinder it, 32 Be5, would be met of course by ..Re8.

<4]> In addition, Black can simply play 27..Nxa2. Similar to line <3]>, after 28 Rxe6 (not 28 Be4? Rf1+) ..Nb4 29 Re7 Rf7 30 Rxf7 Kxf7 (diagram), and Black wins easily.

click for larger view

Now back to White’s 27th. Instead of 27 Rxd6 – her “only chance” – she played 27 Bc4?

click for larger view

Black now could’ve won easily, by simply thrusting forward her central pawns: 27..d5 28 Bd3 (28 Bb3 Rf1+, or 28 Bxc3 Rxc4) ..e5 29 Rb4 (or 29 Rg4) ..e4 and White can resign.

click for larger view

Instead, Black increased her “give-back-material” total by playing 27..Rf1+??, dropping a whole Rook. No doubt this was a DEPRESSING moment for Black!

Then after 28 Bxf1 Nxa2?! (I think Black’s best was ..d5 – am I right?) White played 29 Bc4? This looks SO natural, but I believe the analysis of Black’s correct reply (two paragraph below) proves it’s an error; correct is 29 Rxd6 Rxc2 30 Be5!

click for larger view

...and Black will soon lose at least one of her three extra pawns that she temporarily has for White’s extra piece, when White would clearly be winning.

Now back to the position after White’s 29 Bc4?

click for larger view

Black followed with another error, 29..Kf7? – probably thinking, after 30 Bxa2 Rxc2 (next diagram), that she is now threatening both the Bishop on a2 as well as ..Rc1+. But both players missed that White could then play 31 Ra4!, saving both Bishops (Black can’t check and capture on a1, due to White’s Bxe6+), when White would be up two Bishops for four pawns.

click for larger view

So after 29 Bc4?, correct was 29..d5! when best play seems 30 Bxa2 Rxc2 31 Ra4 Rc1+ 32 Kh2 Rxa1 33 Rxa7 (next diagram) and Black has three pawns for the piece. Somehow I feel Black’s best move here is 33..Kf8 and that Black actually has better chances; am I right?

click for larger view

Even if this endgame is only equal, it is truly amazing – after Black’s blunder on move 21, her inaccuracies on moves 23 & 25, and her second blunder on move 27 – that nonetheless, because of how much material White had previously blundered on moves 13 & 16 (totaling a full Rook) – and with White following up with three smaller slips as well (i.e. missing 24 Rxa4, 27 Rxd6 and 29 Rxd6) – it’s amazing that if Black had made the right choice here on move 29, she would’ve still been “in the game”.

Now back to the play that followed (after Black’s 29..Kf7?) White’s 30 Rf4+. It should’ve been obvious that Black needed to move to some square NOT allowing White to insert Bf6 **with CHECK** – because if White gets that move in, and then Bxa2 and..Rxc2 follow, Black is only threatening the Bishop on a2 and thus remains down in material. With this in view the only move meeting this requirement is 30..Ke8. True, White seems clearly superior in this ending – after either 31 Bxa2 Rxc2 32 Bxe6 Rc1+ 33 Kh2 Rxa1 34 Bg8!, or keeping the Bishop-pair which seems even better, 31 Bxe6 Rxc2 32 Bd4. But this was clearly better than 30..Ke7? after which White checked on f6 and then took on a2, remaining with two rampaging Bishops versus four pawns; after this there were no more surprises.