Im pretty new and i was analyzing the opening stages of one of my recent games. I was playing white and we started off
1. e4 c5 2. Nf3 Nc6 3. d4 e6
[StartPly "6"] [FEN ""] 1. e4 c5 2. Nf3 Nc6 3. d4 e6 4. Nc3 (4.d5 Nd4 5.Nxd4 cxd4 6.Qxd4 Qb6 7.Be3 Qxd4 8.Bxd4 exd5 9.exd5 Ne7 10.Nc3 Nf5 11.O-O-O Nxd4 12.Rxd4 Bc5 13.Re4+ Kd8)
Now my next move was
Nc3 (but doesn't really matter) and the lichess computer analysis marked this as a mistake and proposed playing d5 instead (leading to the computer generated variation
4.d5 Nd4 5.Nxd4 cxd4 6.Qxd4 Qb6 7.Be3 Qxd4 8.Bxd4 exd5 9.exd5 Ne7 10.Nc3 Nf5 11.O-O-O Nxd4 12.Rxd4 Bc5 13.Re4+ Kd8) which presumably has me 1.4 pawns up. I looked up this variation on chesstempo and found that it was never played in 2200+ vs 2200+ play, but used 57 times in all games of the database. In those game it lead to white winning in 82.5% of the games.
Which variations make this move "bad" in top level play (except for the fact that
3. d4 e6 wouldnt happen in the first place aka preselection against this variation)?
What are good alternative 4.?
If the d5 variation is not that strong, why does the computer engine still recommend it. Is it just an error in the way it assesses the position, is there a lack of depth / computing time?