Best defenses against e4
If we can trust recent world championship matches, candidates' tournaments and ICCF games, then it is just:
Open Games
Sicilian

If we can trust recent world championship matches, candidates' tournaments and ICCF games, then it is just:
Open Games
Sicilian
French and Caro-Kann are both better according to engine

I don't think Stockfish is perfect at figuring out the openings, at least not yet. However, out of curiosity to see where the Scandinavian defense I use ranks I ran Stockfish 15 to depth 41. The two moves that most elite players prefer moved to the top. My defense came in a tie for 7th ranking but I'm going to continue playing it.
1...e5 (0.10)
1...c5 (0.28)
1...Nc6 (huh?) (0.32)
1...c6 (0.34)
1...e6 (0.37)
1...d6 (0.46)
1...a6 (0.63)
1...d5 (0.63)
1...Nf6 (0.71)
1...g6 (0.76)

I don't think Stockfish is perfect at figuring out the openings, at least not yet. However, out of curiosity to see where the Scandinavian defense I use ranks I ran Stockfish 15 to depth 41. The two moves that most elite players prefer moved to the top. My defense came in a tie for 7th ranking but I'm going to continue playing it.
1...e5 (0.10)
1...c5 (0.28)
1...Nc6 (huh?) (0.32)
1...c6 (0.34)
1...e6 (0.37)
1...d6 (0.46)
1...a6 (0.63)
1...d5 (0.63)
1...Nf6 (0.71)
1...g6 (0.76)
how to get more than 5 lines
Also i think different computer has different results

1...Nc6 (huh?) (0.32)
It is a much better move than its popularity would suggest. 2.d4 is the only real attempt to punish it, since anything else allows 2...e5 transposition to Open Games, but after 2.d4 d5 it is surprisingly difficult to find any advantage for white.
The reason it is not played often is mainly because 2.Nf3 forces black to either play something suspicious or to transpose back into 1...e5, and if black wanted to play an 1...e5 opening he would have done it on move 1.

If we can trust recent world championship matches, candidates' tournaments and ICCF games, then it is just:
Open Games
Sicilian
French and Caro-Kann are both better according to engine
The French is practically busted at engine level.
Another example of how using engine to evaluate first moves doesn't work

According to stockfish NNUE on depth 22, the best defenses are
French
Caro-kann
Sicilian
Open Games
And St. George's defense (a6)
Apparently every other move is worse than skipping a move.
At depth 22, you are barely at the tip of the iceberg. The horizon effect is going to give you really bizarre results.

According to stockfish NNUE on depth 22, the best defenses are
French
Caro-kann
Sicilian
Open Games
And St. George's defense (a6)
Apparently every other move is worse than skipping a move.
At depth 22, you are barely at the tip of the iceberg. The horizon effect is going to give you really bizarre results.
at depth 44 it still tied with scandinavian

at depth 44 it still tied with scandinavian
You realize that 44-ply is 22 moves. The average game last 40 moves, so you are a little over halfway to the average human game. The average computer game goes well into the 80-100 move range, so there you are about 20-25% of the way to analyzing to that level. As you get further, it will take longer and longer to go deeper. Right now you are looking at the horizon effect and wondering why you are getting funny results.
There is a thread already where someone left the engines running on infinite analysis for several weeks to see what openings the engines liked better. It was a massive waste of electricity that you are trying to duplicate here.
According to stockfish NNUE on depth 22, the best defenses are
French
Caro-kann
Sicilian
Open Games
And St. George's defense (a6)
Apparently every other move is worse than skipping a move.