Why does the lichess engine consider the King's Indian an inaccuracy?

@IAmMateCheckMate , humans doing blunders here and there and getting win/ loss is too far from chess engines accuracy level.

If you look at statistics from Tcec and other high rated correspondence ICCF play, black will have high chance of losing.

Inaccurate play doesnt mean you " lose like blunder" but giving some favours/ winning chance to opponent unnecessarily.

So, what is accurate play?. There are solid known lines for black such as QID, QGD, Nimzo.

The KID does well in correspondence chess and engine matches. Old myths die hard.

...g6?! is due to the very limited horizon.

If you use computers please use them properly!

Forget the computerevaluation. The Kingsindian is equally good. Just study Fischers games.. And Nakamura, I think? I think you need to be a good attacking player..

The King's Indian Defence is one of the rare defences where black won a game in the TCEC engine verus engine superfinal.
Just ignore engine evaluations in the early opening stage. These are inaccuirate due to the complexity of 32 men positions and thus the horizon effect.

@drmrboss High rate of black losing in engine games is because black is going for some pathetic lines which looks stupid for humans. As the @Sarg0n said, if you use engine properly (not in the opening stage) you will see better quality of your play. So I don't know why you are putting the engine on after few moves on the board, put it in the middle game.
If it's so bad why did Nakamura played this for so long and other top GMs?

@IAmMateCheckMate , if you look at database of any human or computer games, KID has higher chance of losing both human and engine games, compared to other solid lines. (database from 365 website)



As chess is a theoretical draw game, even 1. d4 h6?? is still likely a draw. Same to 1. d4 g6 etc

But black require a lot of accurate follow up to prevent losing due to suboptimal play of those first couple of moves.

Of course Stockfish can still make a draw with accurate play after g6?? but you will probably need x10 cores or x10 times to find accurate follow up in crampy position.

2... g6 can still transpose to Gruenfeld.

According to lichess.org/blog/WFvLpiQAACMA8e9D/learn-from-your-mistakes, which is admittedly 4 years old, the Lichess analysis was suppose to avoid marking such moves as mistakes.

I constantly find the Lichess analysis done from "Request a computer analysis", suspect. It marks moves incorrect that are not, and it does not mark moves that are incorrect. I know that last thing was coded that way on purpose to avoid littering the annotations with glyphs in a probable winning position. However, for learning purposes, finding out in post analysis that you hung a piece, even if you were winning, is desired.

Currently, the only solution I see is:
Make a study of your game. Run the analysis. Review your game with a strong player if you can. Review the game and use the master database for evaluation of opening moves, and use stockfish on each move. See if stockfish's score for a move drops from the half-move before by substantial points (say around 3). That could mean a tactical mistake. Use the tablebase when appropriate.

When I do the above, I remove the inappropriate glyphs and comments, and add appropriate ones.

I mean a lot of common moves are labelled ?!, including King‘s gambit, the modern 1. ... g6.

#17, of course
1.d4 h6 can still transpose to QGD, London, or a lot of opening.