Actually, I've been playing the French for 30 years and even I find this one bizarre. I've tried the "Right lads, it's move 14, everyone back home now" approach a few times. I tend to get squashed.

3...h6 is quite respectable after 3.Nd2, when White's slower development means that Black can get away with a waiting move. But 3.Nc3 h6 is much riskier. Berg must be an ice-man.

Totally unbelievable!!!!

Here's the first interesting move: 3...h6:

click for larger view

Black more normally plays a committal move at this point, such as 3...Bb4 (the Winawer) or 3...Nf6 (the classical). But both of those moves allow white to follow a different counter-attacking plan - eg to meet 3...Nf6 with 4. e5 and f4.

So black's 3...h6 is a waiting move. Now it's white's turn and black gets to react. If white plays 4. e5, it doesn't hit a knight on f6 and so doesn't come with tempo. If white plays 4. Nf3 then he is ruling out a an immediate f4 (ie to support e5). The Bc1 doesn't know where it wants to move to next. Moving the Bf1 allows c5 (as in the game) or my favourite b6 followed by Ba6 to exchange off black's bad bishop.

The other popular waiting move for black is 3...Be7 (more usually seen against the Tarrasch). Same principles.

Fast forward. By the time that we get to 9. e5, it turns out that black is now playing the white side of an advance french type position with advanced pawns on d4 and e5.

click for larger view

These pawns cramp white badly. White doesn't have enough free space for all his pieces. In particular his knights have few safe spaces. The Ne2 can only go to g3. The Nf3 has only awkward sideways or backwards movws to make.

And this gives black a simple plan. He needs to avoid the exchange of pieces (as this would ease white's cramped position) and then he needs to throw his pawns forwards to embarrass the immobile white knights.

And that is why black retreats all of his pieces. He wants to avoid piece exchanges to keep white hemmed in. The centre is locked so the loss of time isn't really significant. Black's much greater control of space means that he can happily uncoil from his back rank. White has to operate in a much smaller space - and would really like to exchange off a few pieces.

Fast forward again to the position after 11...Bd6

click for larger view

White's plan is to play Nh4-f5 and force an exchange of minor pieces. So black first hops his knight out of the way (12...Ng8) and then he tucks his bishop safely away (13...Bf8). That way the move Nf5 doesn't fork black's bishop and knight. Throw in a token 14. b5 Nb8 and we get to here:

click for larger view

Black's back rank retreats have trapped the white Nf5. White can do nothing to prevent g6 from winning this knight as it has no safe retreat squares.

All in all, black's idea to retreat everything to the back rank is logical. He wants to avoid exchanges, push white back with his pawns and then redevelop his pieces with interest.

But White was never winning: she was going wrong by move 10 or so, and Black is always better after that. The piece sac is White's best try, but it falls short.

And White is completely lost near the end. Another nice alternate line is 41.Rxd7 instead of Qf3.

Black doesn't recapture, but plays 41...Re8+ and mates. If White tries to protect the Rf2, then 41...Re8+ 42.Kf3 Qd1+ and mate next move.

I like this game so much, though - a proper dissection is required, with silicon assistance and grain of salt.

Evaluating those positions is a whole nother story. Black doesn't get a centralized Queen, which is important, and any exposure of the Black King looks dangerous without counterplay.

Without getting into engine spew, the plausible 20...Nf6 is one of those moves that gets worse the deeper you look. Maybe 20...f6 is playable, but it creates new light square problems and Black is forced to defend.

OTB, I have to agree that 20.Nf4 would offer good winning chances. Much better than that weird Qb1-b3 maneuver, anyhow.