I was always fond of this line for blitz chess. DEspite its reputation there is a big list of famous players who have played it.

I believe both variations with 4. ... Nf6 might be better suited against the two knights system, as the white knight has been committed to f3 already. Jobava used 5. ... exf6 in his famous victory against Karjakin, so it's definitely something. 5. ... gxf6 seems fine there too, at least in a sense that white doesn't achieve a crushing score even if he plays the most precise theory.

I might also point out that my interest here is in correspondence chess. It seems as though it's possible to get into relatively uncharted territory well before engines can provide definitive help to the less familiar player.

HgMan wrote on 09/10/17 at 03:13:46:

https://www.everymanchess.com/dangerous-weapons-the-caro-kann

http://shop.chessbase.com/en/products/davies_the_4_nf6_caro_kann

1.e4 c6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nf6 5.Nxf6+ gxf6

I don't want to pretend that this is superior to the mainline Caro Kann, but is it theoretically sound? And to what extent might one acquire any winning chances with the Black pieces? Database stats look skeptical, but there doesn't seem to be a steady development in the theory.

Negi recommends 6.Nf3 Bg4 7.Be2, but his survey is fairly cursory. Alternatively, 6.c3 Bf5 7.Nf3 seems to score exceptionally well for White. But it seems as though the unorthodox array of pawns can result in some pretty intriguing late middle games and endgames that tend to favour the stronger player. Sound right? Any thoughts? Is there any literature on the Bronstein-Larsen other than Silman's fairly old book?