Openings and computers

I'm sure it's been asked before, but isn't about time opening theory gets rewritten with the help of computers? 

Given that top engines have superseeded top humans, shouldn't we rely on them for best opening moves rather than the analysis of grandmasters from 20, 50 or 100 years ago? 

Surely opening theory wasn't developed by only one person and definitely not during one standard time control game. But again, given the strength of top engines, it seems to me they should get the upper hand. Ten years ago Deep Fritz 10 defeated Kramnik 4-2, yet nowadays isn't even in the top 50 engines. Not to mention the advances in computing power...

I dont play engines in tournaments, i play humans, who play human moves.  I dont need perfect opening theory.

Another issue is whether or not computers produce perfect opening theory. They may defeat humans at common tournament time controls, but does that mean that computers (by themselves) will do better than human-computer collaboration when it comes to home preparation? About three decades ago, the late IM Michael Valvo challenged an early version of Deep Blue to a pair of games where both sides could use 24 hours to choose a move. Michael Valvo won both games.

http://www.chessgames.com/perl/chessgame?gid=1472110

http://www.chessgames.com/perl/chessgame?gid=1472109

I do not remember hearing of a more recent similar experiment.

"... we now have computer-checking of variations and vastly-expanded database resources, ... I also feel that NCO's set of expert and hard-working authors is eminently suited for this task, ...

There are a number of other good features to [Nunn's Chess Openings]. The computer-checking itself should practically eliminate one- or two-move tactical oversights (I haven't found any yet). ..." - IM John Watson (1999)

http://theweekinchess.com/john-watson-reviews/nco-a-preliminary-look

"... What has changed in eight years? In 2003, pc programs were already roughly on a par with the human World Champion, but they achieved this level by being much stronger in tactics whereas they were only at strong amateur level positionally. Now the best programs are far stronger still tactically, but vastly better positionally – maybe in the low grandmaster range, though this is very subjective. ... I do on occasion disagree with the engines (even my own) and substitute my own analysis, but I generally only do so when I am fairly confident that the engines are wrong. ..." - GM Larry Kaufman (2011)

https://www.newinchess.com/Shop/Images/Pdfs/955.pdf

Thanks for the interesting info... I'm still browsing through :)

I wasn't aware chess engines were/are so weak positionally. 'Weak' in this case is a relative and deceptive term, of course. They're only weak positionally compared to their killer tactics. 

If in 2003 engines were on par with strong amateurs and in 2011 they reached 'low grandmaster range' (what does that even mean, 2500-2550?), would it be reasonable to assume by 2016 they have caught up with top humans as far as positional 'understanding'?

With regard to the writing of opening theory, you would want to know whether or not a computer thinking on its own for hours or days could do better than a human-computer team working for hours or days. I have not heard any reports of this. I would think that it would be widely reported if it had been found that the writing of opening theory could be completely automated.

All the top players essentially have super computers running 24/7 ironing out opening lines. This has already been happening for years (and one of the reasons why heavily theoretical lines are often avoided by top players, they don't want the game to be decided by who did better home prep).

Due to how engines work though, they're not good in the opening, so humans have to assist them when exploring openings. Humans lose to engines in man vs machine matches because humans make large blunders more frequently. Engines make noticeable mistakes too, but too small for humans to exploit them to a win during a game.

Some positions have no tactics. That's when even players as weak as me will question their engines... especially if I've only let it think 10 or 20 seconds.

Also sometimes their suggested moves are hugely impractical and based on (of course) enormous calculation. It's impossible for engines to find, for example, high percentage plays so to speak. If an incredibly risky winning line is 1/100th of a pawn better according to the engine's programming, then it will prefer that move no matter how easily other moves win.

I agree with Diakonia, we are not computers, we are humans, a +.50 to me means nothing if I cant see 24 moves deep and the position in front of me looks suspicious.