Why do chess computers often misevaluate the french defense?

It doesn't nessisarilly give Black an immidiate advantage, but I think it is rather quick to claim equality.

And in the Steinitz variation, I actually prefer White's pawn structure as it often proves a nice outpost on d4 for a piece after dxc5 or ...cxd4.

Why do chess computers often misevaluate?

I don't stand in French spesific because the character of the question is not a particular incident.

I believe the truth is that humen made the engines understand and evaluate every position with the same type of equation/algorithm that not match every situation. It's like solving exercises in mixed chapters in physics that involves to calculate dynamics of space-time-mass with several types and instead you use only one and the same all the time.

The humen arrogance to claim that understand chess through engines is such a tragicomedy that not only deny to accept the failure but furthermore publish books ( GM Lary Kaufman) based on engine evaluations which is completely ridiculous IMO, but I understand that everyone wants to make money somehow in this world.

The only real value to estimate opening positions, theoretical tabiyas or not, is through studying all the variations and it dynamic potential till the smoke becomes clear and practical play becomes understandable.

Queens Gambit Declined : It took half a century and more of practical experience for GM's to conclude that in order to play this defense successfully you should prevent e4 first in priority and then concern for anything else. Now if you can find someone to make an engine understand this importancy in the opening phase I recall my statement, but you will not....

In conclusion, engine evaluations for opening phases or closed positions are not facts that worth to take them seriously.

Ummhh..Well huh.. Naka beat a comp in a damm closed position and humilated it by underpromoting 6 Knights something.The french leads to such positions

mabye for stockfish the french is the best because stockfish has 300,000 french games played by GMs in its evaluation function.

   But that dosent mean all the games are gonna be perfect , because theirs always a single winning line that the machine has difficulty pruning , and like i said earlier , engines rob you of creativity  , and without creativity your phucked.    unless you can hardwire an engines material table for misplacing light and dark bishops in ur brain.

   in my opinion an engine wont help u improve much if at all.

  Our brains operate like a computer as it is, each one of us is like an engine in our own unique way....   especially spassky

"I critically look at the computer Houdini, who is a trainer for all. I look at its assessments and evaluations and think they are simply awful! As they say, one's ears burn when one looks at these evaluations! Both in the opening and the endgame too. And after combinations. I can cite hundreds of examples where the assessment is completely wrong, just absolutely! Well, in the opening on the first moves, the computer does not understand anything. Even on the fifth, tenth move, there are a lot of errors. And all the young people think that the computer is infallible. This is not true." - GM Evgeny Sveshnikov

Did he say that this year?  Or awhile back?

For anybody who's read Sveshnikov, a little over exaggeration in expounding his own view shouldn't be surprising.

While computer evaluations do have issues. It would seem most of the time, with a top engine, if one basically takes anything less than .75 as pretty much equal, .75 to 1.0 as noticeably better, 1.0 to 1.4 as likely winning, greater than 1.4 as winning, evals are pretty much right on.

I'd guess most of these evaluations being way off stories, even in endgames, might relate more to weak programs, inadequate computing power, or simply reading too much into the eval.

While its probably true that the program doesn't understand the KID, it's strange that Nakamura probably has near 0% chance of beating a top program with a KID straight up. This seems to suggest that while the program might not understand the KID, it's evaluation of the positions is still pretty good. 

Svidler has got it right. The engines are not flawless and in the opening, since the eval rarely gets beyond equal or say plus or minus .70, the usefulness is minimal.

But beyond the anecdotal GM statements that evaluations are often wrong. I've yet to see many, if any, examples where a good engine significantly misreads a position relative to the parameters in post #30.

The chess program's sole aim is the evaluation of the position. To quote from Deep Blue's summary, "At the heart of Deep Blue's ability to play chess is its evaluation function. The evaluation function is an algorithm that measures the " goodness" of a given chess position...Deep Blue's evaluation function looks at four basic chess values: material, position, King safety and tempo. "

While current programs are significantly more advanced than Deep Blue, the modern programs basic goal for getting to its most accurate evaluation still appears intact.

Now I'm no big fan of computer programs in chess. In fact, I think it's a major detriment. But regardless of Sveshnikov's and some other GM's bluster, the success of programs in play and the lack of a number of meaningful examples make the assertions that engines often wrongly evaluate a position almost ridiculous.

"Chess engines are often detrimental to the chess health of non-masters"

"Don’t expect a computer to help you with anything but tactics." - IM Jeremy Silman

A chess engine is NOT your friend (Hyperlinked article by IM Silman)

Here is what super-GM Boris Gelfand says in his excellent book, Dynamic Decision Making In Chess:

"I suggest that those who start out young should study chess without computer assistance for years, in order to understand the game before you use this powerful tool. As Kasparov said: the main thing to understand about engines is when to turn it on and when to turn it off."

2 interesting articles by GM Gregory Serper:

Should we trust computers?

Do not misuse your chess engine!

I think the quote can be found here

(the page is not loading for me right now, but I think it was this one)

http://chess-news.ru/en/node/19535

There are still many things that computers evaluate wrongly.

Here is a recent engines-on game I have won as white in a Steinitz French. Around move 18 engines evaluate the position as almost equal, while a chess master does know white has the advantage, and has to form an effective plan to utilize it. In the former game white mechanically played all engine recommendations and only drew. I think I did a tad better- white's winning plan is rather complicated, but really cute: The idea of planting the knight at f6 after appropriate manouvering is too much for black (with, or without an engine) to handle...

Computers are not gods. We are humans not computers, and we must evaluate situations as humans. Computers cant think, they can just calculate. Many positions are losing for computers when practically for humans they are winning.

Lord AK47 AKA Sveshnikov said that on 2015.