click for larger view

So, what constitutes White's advantage that allows Reshevsky to win the game? All I can think of is that White's bishops are pointing at Black's k-side pawns and Black's bishop are not attacking anything. This allows White to gain the initiative by keeping Black's pawns under attack while increasing the mobility of his pieces and driving Black to passive defense.

Compare the position just 5 moves later. After 33.Bc5+ White's bishops control the board, Black's king is stuck on the 8th rank, and White can start his h-pawn's advance to wreck Black's k-side position.

click for larger view

By 38.Kxh5 it's over. Black's h-pawn will fall since it can't be defended by either Black's LSB bishop or king and his f-pawn will also fall since it can only be defended by Black's LSB bishop against the combined attack of White's king and LSB. And with a 2 connected passed pawn advantage, it's an easy win.

So in hindsight I think that 30...Bb6, keeping White's king tied to the defense of the Pe3 and preventing Bc3-d4 (which would result in an isolated pawn after ...Bxd4) is better. Conversely, White should have played 29.Bd4 preventing ...Bb6. Therefore Black should first play 28...Bb6 and if 29.Kf2 Bf5 preparing to retreat and protect the Pa6 if necessary. Or he should give up the Pa6 after 28...Bb6 30.Bd5 Bxe3+ 31.Kf1 Bd3+ (to protect the Pb4 after Bd5-b7). Notice the reversal, in this case Black's bishops have the same commanding position in the 3rd rank that White's bishops achieved in the game.

click for larger view

Black's 34...g6 also looks like a mistake, allowing White a target exploited by Reshevsky's temporary pawn sac 35.h5. Black could try 34...h6 to prevent the h-pawn from being picked off by Bd5-g8 followed by 35...Bb8 and 36...Bc7. As long as Black's DSB stays in the b8-h2 diagonal, Black's LSB on c8, and Black's king stays on e8 (to prevent Bc5-f8) it's not clear to me how White makes progress since Bd5-c6+ is simply answered by ...Kf7.

click for larger view

Once again it doesn't seem that Houdini has any idea what to do next and how to make any progress in the endgame so its eval of [+0.04] is justified as far as it is concerned.

Rybka 4.1 evaluated the position after 28.b4 as slightly favorable for White at [+0.26], d=30, also deviating from the game at 30...Kf8 and following Houdini's main line until 36...Ke7. Rybka's main line was: 28...Bf5 29.Bd5 Bc8 30.Kf2 Kf8 31.f4 g6 32.Bd4 f5 33.Ke2 Ke7 34.Bc5+ Kf6 35.Kd3 Bc7 36.Bd4+ Ke7 37.Bc3 Bd8 38.Kd2 Bc7 39.Kd3 Bd8 40.Kd2 Bc7 41.Kd3 Bd8 42.Kd2 Bc7 43.Kd3 Bd8 resulting in the following position:

click for larger view

So Rybka also follows Houdini's useless White king's trek to the q-side where it doesn't accomplish anything. And it's somewhat odd that it evaluates the position at [+0.26] given that it just achieved a draw by repetition. So no kudos to Rybka in this endgame either.