Now it's time to get into some ugly theory. For reference, I'm working off of Gallagher's Starting Out: The Caro-Kann, McDonald's Main Line Caro-Kann, Jovanka Houska's Play the Caro-Kann, and Kasparov's On Modern Chess Part 1: Revolution in the 70's, as well as recent games from This Week in Chess.
This time we'll take a look at Black choosing (or feinting?) to castle queenside. This was originally not only most popular but essentially thought to be forced, as castling kingside was thought to be suicide. Bent Larsen, the tireless pioneer, has changed the evaluation entirely and now kingside castling is thought to give Black the best dynamic chances. Nonetheless, queenside castling is still popular at the amateur level as Black's defenses look very solid. To prove otherwise White must play with some verve.
Starting from our earlier position after 10. QxB:
Black is going to follow with the moves ...e6, ...Ngf6, ...Qc7 and then potentially ....O-O-O. Order doesn't matter much but note that castling is saved for last on first principles of maintaining flexibility - as we'll see he may change his mind.
10. .... Qc7 (note that ...e6 or ...Ngf6 can be met with 11. Bf4, but after ....Qa5+ 12. Bd2 Qc7 Black transposes back to the main line. 12. c3 is thought not to be best because the queen is well posted on a5 and Black can castle kingside safely).
11. Bd2 Ngf6
12. O-O-O e6
And now White has two critical choices - 13. Qe2 and 13. Ne4, with different underlying strategies...
A concept first developed by Spassky and good enough to give him his first win
in his 1966 world championship match versus Petrosian. The idea is simply to place a powerful knight on e5 as quickly as possible.
Spassky patiently explains the first bust to the Caro-Kann
13. ..... O-O-O
14. Ne5 Nb6 (..c5!? will be examined below; ...Nxe5 leads to a pawn formation that is typically poor for Black in either the middlegame or endgame).
A good move which attacks d4, defends f7, and prepares a knight jump to d5.
15. Ba5 Rd5 (15. c4!? is a dangerous gambit, meanwhile after 15. Ba5 c5?! 16. Rh4! Black faces possible problems with a rook bearing on the c-file skewering the queen after 16..... cxd 17. Rdxd4!)
16. BxN axB (16. b4 is exceptionally dangerous after the exchange sac ...Rxa5!)
17. c4 Rd8 (17....Ra5!? is also possible though risky with complications that are analyzed for a further 10 moves in Kasparov's excellent book)
Source games that expand on the position after 17 ....Rd8 include Spassky-Pomar
1968 and Mecking-Pomar
1968, where it was thought that Black had enough resources to hang on for a draw. However, in the modern treatment White does not exchange pieces as early is in the above games, and Black is getting hammered in both 17...Rd8 and 17...Ra5 (65% score for White).
Conclusion: After Spassky's 13. Qe2, 13... O-O-O is virtually busted.
13. .... c5!?
The key move, which required proof that castling kingside can be OK for Black in the Caro-Kann to be discovered. This is where flexibility pays - Black sees 13. Qe2 and decides to change plans, instead asking the question of why Qe2 is a good move?
14. Rh4!? Be7 (note the Rh4 maneuver again)
15. PxP NxP
16. Rd4 O-O
17. Bf4 Qa5
Later Black began to play more ambitiously with 14....Rc8!, when his attack is more advanced than White's even though the Black king is still in the middle. White is not doing as well here in recent games, and 13. Qe2 has taken a back seat to a new concept....
13. Ne4! O-O-O (now ....c5 is just bad after 14. NxN+ NxN 15. Qb5+)
This move was Geller's novel conception - the simple but effective idea is to put the bishop on f4 and knights (if allowed) on e5 or even d6. The resulting line is the modern main line. Notice, incidentally, how Ne4 has reactivated White's worst placed piece. It may seem that this allows Black freeing exchanges, but watch how each of White's remaining pieces arrive at ideal squares before Black's.
14. .... Nxe4 (14.... Nc5 looks cute but after 15. NxN BxN 16. c4! Bd6 17. Bc3 Kb8 18. Qe2 and 19. Ne5 it seems White is having all the fun and is scoring better than 62% recently)
15. Qxe4 Nf6 (an important moment - the knight can also stay on d7, but this allows a strong line from White involving c2-c4 and d4-d5. See Frolov-Volzhin 2000 for details - although the game ends well for Black the opening looks strong for White. Though oddly Gallagher and McDonald come to opposite conclusions on the very same game.)
16. Qe2 Bd6
(16. ...c5!? 17. PxP Bxc5 18. Rh4! was part of a famous bust, Tal-Hubner 1979
17. c4 c5!? (Black waits for the c4 square to be occupied before trying this)
18. Bc3 cxd4
19. Nxd4 a6 (preventing Nb5)
20. Kb1 Rd7
And the battleground has been set for a late middlegame where White looks to play on the c-file and the strongpoint on e5 after, say Nf3-e5 and Rc1. Meanwhile Black looks to play on the d-file. Notice how the pawn on g3 blunts Black's bishop, while the queen on e2 is perfectly placed to defend c4 and h5. White has been doing well here - scoring 60% by my database.
Conclusion: Black is under some pressure in Queenside castling lines. The line is a refuge for endgame sadists. But as the cop told the Joker in Dark Knight, 'I know you're going to enjoy this. I'm just going to have to try to enjoy it even more.'
I'll cut off discussion here because 1. I have a day job and 2. If you ever reach this position, you know you are and are most likely playing a master/Caro-Kann fanatic as you are now 20 moves into theory. That said, I find theory in the mainline Caro-Kann flows like water...one move flows seamlessly into the next....
Next time: kingside castling, the modern way.