Annotated Game #177: How could I not win this?

While it's always disappointing to lose a game, there's another - sometimes just as poignant - feeling of disappointment at not winning a game.  This next tournament game falls into that category.  I build up excellent attacking prospects on the kingside, with open lines and an overwhelming local superiority of pieces (4-0), but at the crucial point I failed to actually execute an attack.  My opponent started to do a good job of defending while making threats and turned the game around as a result; what happens after move 27 is an excellent illustration of the importance of the initiative, both on the board and psychologically.  I almost had the full disappointment of losing the game, as things went rapidly downhill, but after an error by my opponent I managed to calculate the drawing sequence and wrapped the game up.

Analyzing this game was helpful in highlighting certain clusters of turning points and strategic choices, for example around moves 17-19 and again on moves 27-31.  Hopefully I can make better future decisions as a result.

[Event "?"] [Site "?"] [Date "????.??.??"] [Round "?"] [White "ChessAdmin"] [Black "Class C"] [Result "1/2-1/2"] [ECO "A28"] [Annotator "ChessAdmin/Komodo 10"] [PlyCount "81"] [EventType "simul"] [EventRounds "5"] 1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. e3 d5 5. cxd5 Nxd5 6. Bb5 Nxc3 7. bxc3 Bd7 8. d4 Bd6 9. O-O O-O 10. Be2 Re8 11. Qc2 e4 12. Nd2 {now e4 becomes an easy target to focus on, although the position is still equal.} Re6 $2 13. f4 { here I believed my opponent a little too much, when he failed to defend e4, although the text move is also good for White.} (13. Nxe4 {I correctly saw there was a threat for Black involving ...Qh4, which made me avoid this line. However the engine shows that this is unnecessary, since Black's attack just looks scary rather than being effective.} Qh4 (13... Bxh2+ 14. Kxh2 Qh4+ 15. Kg1 Qxe4 16. Qxe4 Rxe4 17. f3 Ree8 18. Kf2 $16) 14. f4 Rh6 15. h3 Rg6 16. Bf3 Bxh3 17. Nxd6 cxd6 $16 {and White is fine.}) 13... exf3 14. Nxf3 $16 {here Komodo considers White to be about the equivalent of a pawn ahead. The central pawns are quite powerful and I have some nice open lines on the kingside.} Re8 15. Bd3 h6 16. e4 {mobilizing the central pawn majority} Bg4 17. e5 {this is premature; I should bring more pieces into play first, or (as shown by the engine) challenge the Bg4, which is Black's only good piece on the kingside at the moment.} (17. h3 {would help get the pesky bishop out of the way. For example} Bh5 18. e5 {now this pawn advance has more bite, since g4 is threatened and the bishop has no good square to go to. So} Bxf3 19. Rxf3 $18 { and White is rolling on the kingside with four pieces (queen, two bishops and rook) currently versus zero for Black.}) (17. Rb1 $5 {is a simple move that gets the rook in play and makes life more awkward for Black.}) 17... Be7 $2 { an obvious retreat but not a good one.} (17... Bf8) 18. Rb1 {a good move but not great.} (18. Bc4 $1 $18 {pinning the f-pawn causes major problems for Black, as now Qg6 is threatened.}) 18... Rb8 19. Be4 {right piece to move, but not the best square for it (per above). I thought for a while here on where the best spot for the bishop would be, but ultimately was too focused on queenside play, when in fact the big payoff is on the kingside.} Qd7 20. h3 Bxf3 (20... Be6 {is more solid.}) 21. Rxf3 $18 {with the f-file now open and two strong bishops pointing towards the king, along with the dominating central pawns, I have a major advantage.} Bd8 22. Rg3 {going for the somewhat cheap-looking, but still effective, threat of Bxh6.} (22. Qf2 $5) 22... Nxd4 23. Qd3 {seeking to avoid having to take on d4 with the pawn and then give Black ...Qxd4+. However, that would in fact be fine for White as well.} (23. cxd4 Qxd4+ 24. Kh1 Rxe5 {here I was too materialistic and thought that the three pawns for a piece wasn't a good deal for me.} 25. Bh7+ Kh8 26. Bb2 $18) 23... Rxe5 24. cxd4 Rh5 25. Bb2 {lining up against g7.} (25. Rxb7 $5 {would pursue a simpler winning strategy, based on my material advantage.} Rxb7 26. Bxb7 Bf6 27. d5 $18) 25... Bg5 26. Rf1 {now all pieces are in on the attack and the engine evaluates this as the equivalent of White having an extra piece and then some.} Re8 27. Qf3 $6 {a significant slip, since it would be much better to double the rooks on the f-file rather than leading with the queen. Also, the lack of a battery on the b1-h7 diagonal leaves me with fewer options for the bishop and allows Black's next move. I did not in fact have a concrete plan here.} (27. Rgf3 Bf6 28. Rxf6 $1 gxf6 29. d5 $18 {and Black's king is stripped of cover.}) 27... Rh4 {well played. Black now starts taking back some of the initiative by making his own threats, the first in a while.} 28. Bf5 Qb5 29. Qc3 {now I am responding more to Black's threats than looking for my own.} (29. Bd3 {would take advantage of the weakness on f7.} Qd7 30. Qxb7 $18) 29... Rf4 30. Rxf4 Bxf4 31. Rf3 $2 {missing the way to keep an advantage. The position is now equal.} (31. Bd3 $1 $18 {extracts the bishop with tempo, also saving the Rg3.}) 31... Qxf5 {by this point the game has fully turned around and Black is the one with all the threats. Psychologically this was a blow and I was tired of calculating variations, prompting the next error.} 32. g3 $4 { this in fact should now lose.} (32. Bc1 {is in fact the only move to preserve the draw, as it blocks Black's next.}) 32... Qb1+ 33. Rf1 Qxa2 $2 {allowing me to draw.} (33... Be3+ {should win for Black.}) 34. Rxf4 {at least I was able to correctly calculate the next sequence.} Re2 35. Rf2 Qb1+ 36. Bc1 Rxf2 37. Kxf2 $11 {now the engine evaluates the position as dead drawn.} Qf5+ 38. Kg2 Qe4+ 39. Kf2 Qf5+ 40. Kg2 Qe4+ 41. Kf2 1/2-1/2