King's Indian Defense Classic game: Theory at odds with computer.

   In this game, Black classically mates White in the King's Indian Defense. The game seems to be how not to play - how to get mated in the middle game. Yet, Chess.com's computer lacks knowledge of thematic KID play (check move list) and even calls several of the key moves blunders!

Enjoy, this is a game every King's Indian player should know by heart, the strategies in it are fundamental.

p.s. I have also posted an article of this, to make sure more people get to read it, it is a good lesson for both white and black.

Thank you shuttlechess, very nice lesson!

So fast so good.

Yeah, one of the not so secret secrets to winning in ICCF is to find openings like the KID where the engine evaluations are so far off-base.

While many fans of engines will claim the evaluations are correct, the engines still can't quite come to grips with the nature of Black's K-side counterplay. Left to its own devices, very few engines will play f5, f4 and g5 in the main lines of the KID.

There are a lot of cases where computers think the best moves are blunders.  Perhaps its a question of humans not being able to calculate 20 half-moves ahead in their minds, but the move they know is good is still a winning move.

In the Vienna Opening, the computer claimed my early f4 - the main idea behind the Vienna Opening - was a blunder whereas a book by Russian GM's says, in the same position about f4: "virtually forced!"

In a game I won here, the move where I gave my King The Opposition in a King & Pawn endgame, assuring myself of the win, was called a blunder by Stockfish 8, 20-ply.

They can play anything to the best of their programming.

White allowed you too much space on the kingside

Engines tend to value "guaranteed" space over potential space. in defenses like the KID and in some variations of the french, getting "the second wave" of the territory expansion gives you an open hand in what to do with said space (for example, get a favorable rook file or knight outpost or just strangle your opponent on said side, to make a passer in the third rank etc).

Its also why engines engines dont understand defenses that are deemed fine like the czech benoni. While the potential for a big squeeze is less, the flexibility of the counterpunches is underestimated. the engine has to brute force search ALOT to reduce such openings to a fair evaluations. Humans from Experience and intuition know there is more poison to them, then it first appears.

There may also be an element of, "it doesnt matter if its a 0.3 or 0.6, if there is no clear way to make progress in the position, eval is pointless". You ever faced a very strong player who plays the philidor defense? you play for 20+ moves good moves that the engine says at least let you keep your precious 0.5 advantage but you dont seem to be getting anywhere with it? (until you eventually lose your patience, and the eval then goes from .5 to .25 to woops you are slightly worse?). This may be why those +0.6 or so you see in KID positions may be not too informative.

@darkunorthodox88 - Nice explanation.

Also want to add that computers also assume that the opponent will be making the best move possible and evaluate the score against those sequence of moves. So with perfect play, those moves are indeed weaker and it is looking for the best possible move. However in reality, people tend to make sub-optimal moves, especially under pressure.

Don't we all knowingly make weaker moves against much lower ranked players cause you know they will fall for our trap but wouldn't dare try it against higher ranked players and certainly not programs.

Wow this game got dug up from a long time ago... and my old profile account! Nostalgic...

but that's the thing! engine evals dont tell you how easy or difficult it is to keep a position at its given objective eval. Some 0.6 feel like you are walking on thin ice and the slighest error will lead to your position of control, whereas other 0.6 are borderline comfy for the player because the position has clear plans and no easy way to breakthrough it.

This is why when i teach people how to use engines to make opening choices, i advise them not to pick anything that will evaluate higher than 0.5 for black in its even its most critical line, UNLESS you have very good reason to believe the engine is missing some dynamic compensation and not even strong  prepared human player can exploit (this is a good criterion to have in picking openings in general, pick stuff that could potentially last your whole chess career). IF they think its an exception, i tell them to raise the eval window to about 0.7. more than that, and you are playing with fire

The thing is that the engine tells you  who is wining, not the difficulty of the position. You can have a +0.5 edge but only 1 move to keep it, any other move gives an edge to black. Also you can have and equal position but you have many moves to make and your opponent only have 1 or 2 tricky ones.

The engine is telling you who has the edge if you only play best moves so it's just telling you who was more wining chances statistically becouse nobody plays perfect.

Since I wrote my comment in 2012, engines have gotten so much stronger that it just doesn’t matter what opening is played.

I wouldn´t trust in chess.com engine quick analysis, it´s not very accurate