Deep Fritz 10--Kramnik, 2006, game 6 contained these moves:  18...Nf6-g8 19 Nc3-b1.  That reminds me of the Classical/Winawer Retreat Var in the French mentioned by Neil McDonald in June 2006.  I am not a subscriber; I dont know what analysis he provided but I would like to present mine.  The retreat var from the classical order is 1 e4 e6 2 d4 d5 3 Nc3 Nf6 4 e5 Ng8.  From the Winawer order it is 1 e4 e6 2 d4 d5 3 Nc3 Bb4 4 e5 Bf8 reaching the same position.  At first it seems the retreat is ridiculous but lets give it a chance.  If a player employs the Retreat Var he asserts that white must also retreat with 5 Nb1.  That would transpose to the Advance Variation.  Why would white need to retreat 5 Nb1?  White would need to retreat with 5 Nb1 if the knight were poorly placed on c3.  Why would the knight be poorly placed on c3?  The knight would be poorly placed on c3 because a pawn needs the c3 square after black plays ...c5.  In the Advance Var black does exactly this.  Black plays ...c5 and white plays c3.  If a game transposed to the Advance Var white's move has justification.  The question is can white play 5 Nce2 with good effect?  I believe that white can play 5 Nce2 with a better than usual positional advantage.  Thus black should not play the retreat variation.  However, since white won the Deep Fritz 10--Kramnik game, an idea like Nc3-b1 in reaction to ...Nf6-g8 does hold water!  The conclusion was obvious from the beginning but I had fun examining the idea.