I was replaying the game Nowak - Pachman (Solingen 1968) with an engine running. It found what I thought was an interesting tactic. Here’s the score of the game up to the critical move:
Hmm. I might be inclined to try 33. g4 fxg4 34. e4 gxf3 35. gxf3 creating a passer and undoubling the pawns. Other variations after 33. g4 might at least undouble the pawns and give white a 3-on-2 on the kingside.
I looked at that and figured the Black h-pawn and extra Black c-pawn were more dangerous than White’s e-pawn. There is the possibility of 35 … Be6 36 Bxc4 Bxc4 37 b4+ Kb5 38 a4+ Kxa4 39 Kxc4 h5 40 e5 and White queens first, but 37 … Kd6 38 Kxc4 h5 39 Kd3 Ke5 40 Ke3 g5 41 Kf2 Kf4 42 Kg2 g4 and Black is closer to the queen side once all of the king side pawns disappear.
First of all, 33 g4?? looks like an awful move, which probably loses by force. How can giving up a pawn be the right plan? In the line given, after 35 gxf3, Black just retreats the bishop to e6 and the White pawns are going nowhere. Black will follow with …b5, …a5, …b4+ and then …Kd4. I didn’t look further but that sure looks winning at first glance.
I think Jeff has the right idea. The line he gives after 38 a4 looks like it should probably be a draw. There are some very tricky lines if Black tries to chase the White pawns with …Ke3-f2. I haven’t worked it all out but I have the sense that it’s probably a bad idea for Black to do that (though that might still draw in the end). Most likely, the safest plan for him is just to leave his king where it is and eliminate all kingside pawn moves. Then (I think), he can get one of those defenses where the WK is trapped by the BK in front of a passed a-pawn.
[EDIT: After looking a little further, I don’t think there is any such trapping defense as I suggested. I still think it should be a draw after 38 a4 but working it out is going to take more time than I have right now.]
Then 33 … b5, 34 Bf3 a5 with the idea of b4. White’s bishop is not breaking through, the kingside no longer has a viable 4vs3 threat and the queenside does have a viable 3vs2 → 2vs1 threat.
Example: 35 g4 b4+ 36 axb4 axb4+ 37 Kd2 Bxg2 38 gxf5 gxf5 39 Bh4 Bd5 and the Black king can go to the kingside while the White king has an interesting time keeping an eye on the queenside. White might draw.
If white wants a draw in this position, 1. Bxc4 temporary sacrifice appears to achieve this goal fairly quickly. After all trades, white regains material equality and his queen side pawns are closer to promotion than black pawns g6 and h7.
Bxc4 Bxc4 2. b4+ Kd5 3. e4+ fxe 4. fxe4+ Kxe4 5. Kxc4.
Black has to be careful not to lose this endgame.
… a6! 6. a4 Ke5 7.a5 bxa5 8. bxa5 Kd6, black should draw it with accurate play.
Jeff’s suggestion is what I had in mind. I found the tactic interesting because the initial position doesn’t appear (at first glance) to contain a pseudo-sacrifice. Also, I found pleasing the symmetry after 1. Bxc4 Bxc4 3. b4+ and either 3. … Kb5 or 3. … Kd5 with the pawn push to deflect the Black King.
Both Black King moves lead to a draw with best play. As noted above, some care is required. For example, after 3. … Kb5 4. a4 Kxa4 5. Kxc4, Black has only one move (5. … b5) to draw. I checked this analysis using both Houdini and Stockfish, assisted by Chessbase’s six piece tablebase.
Neither player apparently saw 33. Bxc4. Also, Reshevsky annotated the game on p. 100ff of the March 1969 issue of Chess Life. He analyzed 33. a4, but concluded that it lost for White. No mention of 33. Bxc4.