I know the Levenfish attack is not the best variation to play against the Dragon, but I like to try it in this correspondence game. My opponent played 6...Bg7 which isn't supposed to be good for black, but still I found it a tough game to play. After 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. f4 Bg7 7. e5 dxe5 8. fxe5 Ng4 9. Bb5+ Nc6 10. Nxc6 Qxd1+ 11. Nxd1 a6 12.Ba4 Bd7 13. h3 Nh6 14. Nxe7 Bxa4 15. Nd5 Rd8 16. c4 {last book move, it is clear that both of us used a database until this point} Bxd1 17.Bxh6 Bxh6 18. Rxd1 O-O 19. O-O Bg5 20. b4 Kg7
my computer engine suggested 21. Kf2 +/-, which has puzzled me. Does it mean that white has a winning endgame? Further computer analysis doesn't make any sense to me. Can someone tell the general plans for white after 21.Kf2? I ended up winning anyway in a drawish, but tough endgame where both of us made some mistakes, but luckily my opponent made the last one. Here is the full game:

[Event "Dermans Dragon Championship II"]
[Site "chesshere.com"]
[Date "2008.07.21"]
[Round "?"]
[White "Yenman"]
[Black "dutchdonkey"]
[Result "1-0"]
[ECO "B71"]
[WhiteElo "1843"]
[BlackElo "1687"]
[PlyCount "93"]
[EventDate "2008.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. f4 Bg7 7. e5 dxe5 8.
fxe5 Ng4 9. Bb5+ Nc6 10. Nxc6 Qxd1+ 11. Nxd1 a6 12. Ba4 Bd7 13. h3 Nh6 14. Nxe7
Bxa4 15. Nd5 Rd8 16. c4 Bxd1 17. Bxh6 Bxh6 18. Rxd1 O-O 19. O-O Bg5 20. b4 Kg7
21. Rf3 Rfe8 22. Rdf1 Rd7 23. Re1 Re6 24. Nf4 Bxf4 25. Rxf4 Rde7 26. c5 Rxe5
27. Rxe5 Rxe5 28. Kf2 f5 29. Rd4 Kf6 30. g4 Re4 31. Rd7 fxg4 32. hxg4 Rxg4 33.
Rxb7 h5 34. c6 Rc4 35. c7 Ke6 36. c8=Q+ Rxc8 37. Rb6+ Kf5 38. Rxa6 h4 39. b5
Rc2+ 40. Kg1 g5 41. a4 Kg4 42. Rc6 Ra2 43. Rc4+ Kg3 44. Rc3+ Kf4 45. Rb3 g4 46.
b6 Rxa4 47. b7 1-0

Analysis according to the engine Stockfish 1.5.1 (which I got for free off of chess.com.)

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 g6 6.f4 Bg7 7.e5 dxe5 8.fxe5 Ng4 9.Bb5+ Nc6 10.Nxc6 Qxd1+ 11.Nxd1 a6 12.Ba4 Bd7 13.h3 Nh6 14.Nxe7 Bxa4 15.Nd5 Rd8 16.c4 Bxd1 17.Bxh6 Bxh6 18.Rxd1 O-O 19.O-O Bg5 20.b4 Kg7 21.Rf3
( { [Stockfish 1.5.1 JA] 16:+1.73} 21.g3 Rc8 22.c5 Rcd8 23.Kg2 Bc1 24.Rf6 Ba3 25.Rd3 a5 26.Rb6 axb4 27.Rxb7 b3 28.c6 Rxd5 29.Rxd5 bxa2 )
( { [Stockfish 1.5.1 JA] 16:+1.65} 21.Kf2 Rc8 22.Rd4 a5 23.h4 Bd8 24.a3 Re8 25.Re1 Bc7 26.Nxc7 Rxc7 27.Kf3 b6 28.bxa5 bxa5 29.Ke4 )
21...Rfe8
( { [Stockfish 1.5.1 JA] 17:+1.13} 21...b5 22.c5 Rfe8 23.Rfd3 Rxe5 24.Kf2 h5 25.c6 Rc8 26.c7 f5 27.Kf3 Kf7 28.Ra3 Re6 29.g3 h4 30.g4 Bf6 )
( { [Stockfish 1.5.1 JA] 17:+1.33} 21...Rfe8 22.Rdf1 Rd7 23.Re1 Red8 24.a3 b5 25.Nb6 Rd1 26.Rxd1 Rxd1+ 27.Kf2 bxc4 28.Nxc4 Rd8 29.Ke2 Rd4 30.Rc3 Re4+ 31.Kf2 )
22.Rdf1 Rd7 23.Re1 Re6 24.Nf4 Bxf4 25.Rxf4 Rde7 26.c5 Rxe5 27.Rxe5 Rxe5 28.Kf2 f5 29.Rd4 Kf6 30.g4 Re4 31.Rd7 fxg4 32.hxg4 Rxg4 33.Rxb7 h5 34.c6 Rc4 35.c7 Ke6 36.c8=Q+ Rxc8 37.Rb6+ Kf5 38.Rxa6 h4 39.b5 Rc2+ 40.Kg1 g5 41.a4 Kg4 42.Rc6 Ra2 43.Rc4+ Kg3 44.Rc3+ Kf4 45.Rb3 g4 46.b6 Rxa4 47.b7 1-0

Basically, it said that 21. g3 was the best move, followed closely by Kf3, which would have given you about a pawn and a half advantage.
Your move, however, Rf3, narrowed to about a pawn advantage.

Yay for computer analysis.

I could give a computer analysis of the whole game if you want me to.

Thanks, but I was hoping for some human input. I think this is the kind of position where engine analysis is not sufficient, their event horizon is too limited to calculate everything through up to the endgame. It doesn't seem that white can create a passed pawn. That isolated e-pawn can easily be blocked or exchanged, so I don't see why white would have the advantage after move 20. I even think that e-pawn is a weakness, for it can be pressurised by both rooks and the bishop. White has an extra pawn on the queenside though, so I think that's where white's advantage is. However, I don't think it is easy to create a passed pawn on the queenside either, since 3 vs 2 pawns is a lot harder than 2 vs 1 pawn, not to mention the extra pieces both sides still have on the board. White's knight is ideally positioned on d5, but I don't see how to exploit it while black's bishop is not a bad piece either, it has a lot of maneuvering space and can attack the pawns on dark squares.

In that case:

I would say that the point of Kf2 would be to prevent Black from getting after the e5 with Bf4. g3 also does the trick, but I like Kf2 as it is always a good idea to get the king activated in the endgame, even when there are still four rooks and a pair of minor pieces on the board.

I do see the point of Rf3 though, preparing to double the rooks and push some pawns.

Hope this helps.