Lichess analysis annotated 4. Nf3?? as a blunder preferring (4. Nc3) (see on Lichess). I am trying to understand the ideas that make this a blunder. At depth 23 with Stockfish 13+ NNEU, both were evaluated at +2.1.

If you push the depth to 28, it begins to show a 0.1 preference for Nc3 over Nf3, but I still don't see how this evaluation disparity resulted in the blunder annotation.

My thoughts:

  • 4. Nf3 continues development for White. Castling short becomes an option. Black has no immediate threat with the lone queen and must respond by moving the queen. This gives White the option to develop yet another piece. Black may fall behind in development.
    1. Nc3 also develops a piece but adds a protector to the e4 pawn which may come under attack after ...d5 .

I see no ideas that lead to a strong preference for one over the other. Am I missing some ideas in this position?

I realize that +/- 0.1 in an evaluation doesn't really mean a whole lot, but I am genuinely curious to learn if I'm missing an idea in the position.

At depth 48, 4.Nf3 and 4.Nc3 are both top moves with equal evaluations

Edit: A commenter remarked "I don't believe it did." I present a screenshot below proving this did occur.

Lichess annotates 4. Nf3?? as a blunder.

[Title "White–Black (10 min) | ECO: B12 Caro-Kann Defense"]
[FEN ""]
[startply "6"]

1. e4 c6 2. d4 e6 3. Bd3 Qh4 4. Nf3 Qg4 5. O-O d5 6. Re1 Nf6 7. Nbd2 e5 8. exd5 Nxd5 9. Nxe5 (9. h3 Qe6 10. Rxe5) 9... Qg5 10. Nef3+ Qe7 11. Rxe7+ Bxe7 12. Qe1 O-O 13. Nc4 b5 14. Nce5 Nb4 15. Be4 Bb7 16. Qe3 Nd5 17. Bxh7+ Kh8 18. Qd2 Kxh7 19. Ng5+ Kg8 20. Qd3 Nd7 (20... Bxg5 21. Bxg5) 21. Nxd7 (21. Qh7#) 21... g6 22. Qh3 Nf6 (22... Bxg5) 23. Nxf8 (23. Qh6 Rfd8 24. Nxf6+ Bxf6 25. Qh7+ Kf8 26. Qxf7#) (23. Nxf6+ Kg7 24. Ngh7 g5 25. Bxg5 Kg6 26. g4 Kg7 27. Qh6+ Kh8 28. Nxf8#) 23... Bxf8 24. b3 Re8 25. Bb2 Nh5 26. d5 cxd5 27. c4 bxc4 28. Qc3 Ng7 29. Re1 Rc8 30. bxc4 (30. Qh3 Nh5 31. Qd7 Be7 32. Qxe7 Rf8 33. Qxf8+ Kxf8 34. Nh7+ Kg8 35. Re8+ Kxh7 36. Rh8#) 30... Rxc4 31. Qd3 Rb4 32. Rb1 (32. Qh3) 32... d4 33. Ne4 33... Ra4 33. Ne4 (33. Qh3) 33... Ra4 34. Ra1 (34. Nf6+) 34... Ba6 35. Nf6+ Kh8 36. Qh3+ Nh5 37. Nxh5 gxh5 38. Qxh5+
  • I hope I've met the standard for expectations of askers...if not, please comment and I will improve question to meet the standard. Commented Jun 19, 2021 at 16:58
  • 1
    My Stockfish 11 likes 4) Nf3 the best after a 90 seconds of grinding. lichess might consider Qh4 to be a blunder if its "what is a blunder" threshold is low.
    – Tony Ennis
    Commented Jun 19, 2021 at 18:38
  • 3
    Please check at which depth the annotation has been calculated and evaluated. It might be that these ??s come from a very shallow analysis, and when you increase the depth, the evaluation changes of course, but the ?? signs remain.
    – sleepy
    Commented Jun 20, 2021 at 9:49
  • 1
    My Lichess Stockfish is giving 4. Nf3 as the best move at depth 41, with 4. Nc3 second and 4. Qe2 third. In blitz I'd play 4. Nf3 without thinking twice, since it gains a tempo and so is "free".
    – Allure
    Commented Jun 20, 2021 at 12:17
  • 1
    @sleepy When I run these same moves through Lichess's automatic computer analysis it gives me the same result as OP; it calls the move a blunder. I don't know what the depth of that automatic analysis is, though.
    – D M
    Commented Jun 20, 2021 at 17:34

3 Answers 3


The "blunder" annotation is derived from server-side analysis using a low engine depth, which is usually inaccurate.


Its evaluation drops form +2.3 to +0.0, so it considers it a blunder. Of course this does not match what a human understands as "blunder", but computers are incapable of identifying that.

Your move is a perfectly reasonable one and if that's the biggest type of "blunder" you make in your games, you're on your way to a Grandmaster title.

  • 2
    But why does the evaluation drop to 0.0? White has clearly superior development, if nothing else. Something strange is going on here.
    – D M
    Commented Jun 20, 2021 at 17:35
  • It is possible Nc6 leads to some (shallow analysis) line where Black comes under sustained pressure instead of an immediate Nf3 which can be partly nullified by Qd8 and a closed center where the two tempo disadvantage can be closed quickly. Lichess often marks blunders when it sees "shorter" variants - e.g a mate in 6 "blunder" when a more obscure mate in 5 is possible.
    – Fox
    Commented Jun 21, 2021 at 1:50
  • 1
    Wanting to accept this answer but based on this comment, this answer wouldn't be the most useful to future readers in its current form. Commented Jun 23, 2021 at 12:56

Obviously 4.Nf3 is far from being a blunder, maybe the best move according to stockfish. You are missing ablsoutely nothing here. Generally these "annotations" should not be taken very seriously, helpful only for absolute starters. I personally dislike them strongly, there are just adding noise to valuable engine output.

Curiously in the early days of chess programming it was extremly seriously believed, that it might be possible to "teach" a CPU to play chess, the same way as a human, who at that time was considered the ultimate high point of intelligence. The chess program Pioneer (https://www.chessprogramming.org/Pioneer) is a prominent example. If you would have such a program, it could be likely modified to produce meaningful explanations. AFAIK unfortunately none of these attempts was even remotely successful so far.

Hence if someone presents you human like annotations without further notice, you can safely assume these are very ad hoc and shallow made, since otherwise it would be some kind of breakthrough and the creators would not miss to point that out.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.