fredthebear: <samvega> I generally agree with your comments to this active, entertaining game.
21.Nb6+ is a useful intermezzo check as White has two pieces en prise, meaning he will lose one anyway so try to get some value before being removed. Giving up the White knight w/tempo eventually leads to the Black queen surrendering herself to the White rook instead of the previous threat of queen takes rook w/check, so it had quite an impact benefitting White. (If 21...axNb6 instead, the intermezzo 22.Qxb7+ is handy for the White queen and then rescue the cornered White rook.) Looks like 21...Kc7 would have worked out better for Black.
Yes, 22.Qf8 is better. It would have avoided future congestion, which turns out OK for White as well but was much less clear at the time.
22...Bf8 and Black appears better. The simple 23.c3 is an innocent game changer that many of us would have overlooked.
23...dxc3 was too hopeful and ends up costing Black his queen. It seems 23...Kc7 was better.
26...Bd5 sure looks odd, so White scrutinized it's intentions. Black's offer is an attempt to deflect the White bishop from the diagonal to allow Black's c-pawn to promote.
27...e6 does have a slight defensive purpose but is overly cautious. 27...Nb4 was better. Time is of the essence in this position.
Yes 28.a3 seems best. My guess is White simply did not anticipate the Black knight's advance. But the a2 and c3 pawns remain in place for the remainder of the game so it works out to White's satisfaction.
This is one of those kinds of middle games where matters are not quite as seriously threatening as they first seem. Black's responses just don't have enough time and force to fully materialize. White is rewarded for continually increasing the mobility of his pieces to penetrate, removing both Black rooks, and trying to bust the White queen out of jail. Black is simply on the defensive too long in maintaining the restriction.
Then the White queen must kamikaze out of the corner to her death and the endgame has suddenly arisen with White finding itself having the advantage after careful defensive measures. White recoils using a blockade and simplification as the best defense against promotion w/an otherwise won position on the opposite wing.
The opposite colored bishops ending is usually drawish, but White's connected passers on the kingside would dictate the finish. Had the ending played out, White can sacrifice his remaining bishop if necessary (the blockading Bd2 has no worries at the moment) to stop the last of the separated, isolated Black pawns on the queenside. Whereas a preventive sacrifice by the Black bishop only helps the advance of the remaining White pawn as it recaptures on the kingside. Black's monarch won't have enough time to get to hot spots on each side of the board. The OCB ending is better for White this time.