What do super GMs see?

wik, why would it be okay to evaluate it in this way. there seems to be a hypocrisy in the computers play. on the one hand it relies on its computational power on the other hand it admits that it will never calculate anything satisfactorily so falls back to these positional elements. on the one hand it calculates material on the other hand it supplements it with how many legal moves the opponent has or whether it has allowed the opponents bishop to get freed. it should rely on one or the other otherwise, the computer seems to be admitting to its own weakness.

I think you are all correct... GMs have such a large data base of positions to elect from that they can spot the 4best candidate moves almost intuitively and then work backwards from that. Of course they are not perfect but a GM can chose the best moves 80% of the time, while a sub-1800 player would be lucky to chose the best moves 50% of the time.

Kasparove summed it up hilariously when he said... "when a GM looks at a position, he sees his old girlfriends face, an average player sees only her nose hairs" I am sure in Russian it comes out much better but you get the point.

Computers, though, calculate foward (like most average players) the only difference being that the machine has a perfect foward calculating ability and perfect data base to compare with.

GMs know a lot about:

a) To what depth should one calculate? Sometimes there is a need to look many moves ahead. Sometimes 2-3 would suffice.

b) What candidate moves to consider? In other words, a GM sees more opportunities for both sides than a weaker player.

c) How to evaluate the positions being analyzed? Often the problem is not to calculate a line accurately, but to understand who benefits from this transformation - theoretically and practically. At top level most games are lost due to misevaluation, not miscalculation.

d) They calculate faster and have a large bank of well-known positions to which they can refer when analyzing. When a regular GM starts thinking about how to convert an advantage in the endgame, a titan like Kramnik says that it's pretty much over already and moves on

pdve, i think the problem is that there is no way to program a computer to think, to play organically. we have not invented true AI yet.  computers cannot think for themselves. they can compute very quickly, crunch billions of numbers per second, but they can never have an "idea."  given this fact, without a credible (artifical) replication of the way humans evaluate positions, computers would be extremely weak.  the examples that you give, "how many legal moves are available to the opponent, whether a move will free his bishop" etc, these things are coded into the eval function and given a numerical value.  the advantage of computers is that they can assess these numbers in high volume and with high speed, finding the "objectively" best move solely through numbers.  i think this is good, and really indicates the potential of computers. i do not see the hypocrisy that you feel is evident. however, as i said previously, the addition of centuries of human opening preparation and endgame tablebases into the computer are a crutch and should be removed.

also thanks to Natalia for her insight, it is always nice to see titled players engaging with the community here on chess.com

Interesting point from Natalia... something I had never really occured to me... calculation vs. evaluation...

calculation = how (what are the transactions require to arrive at a given position?)

evaluation = why (after these transactions do I have an advantage or not)

I hope I "evaluated" your article correctly.

I also saw an example of this when Gata Kamsky 2750 played a young upcoming IM called Kayden in Round 2 of the US chanmpionships a few days ago. The young IM played (seemingly) brilliant tactical middlegame (at least to us it seemed brilliant) and gained a rook for a knight (seemingly advantage), but Kamsky had played such a superior positional game that the kid resigned 6 moves later early in the endgame, because Kamsky was going to steam roll 3 pawns down the king side.

Of course non of this made sense to us until the final moment and we are wow! And thats why I like to see more open tournaments where 2000players have the chance to play 2700GMs because they are little more understandable. The tournament in Norway (for example) games between super GMs are way above the understanding of most us sub-1800 players, like the Carlsen Anand game was way over my head.

unfortunately, natalia's observations are not really relevant here since she is already a GM and so is used to many of the thought processes which we are discussing here. so she might be taking many things for granted which we find strange and spectacular.

what ?

i mean it would be useful if a GM could actually give an inside view of what it is like to play chess.

I think you mean her observations are the most relevant, since she better than any of us actually knows how a GM thinks. Who do you believe more? Someone who speculates about an answer without having ever actually experienced it, or someone who has experienced it, and in fact worked very hard for it?

Hint: if you believe someone who's only speculating, no one can help you.

there are many arguments that writers make the worst literary critics.

No one calculates 50 moves ahead lol.

no one does but they can

GMs are not an extraterrestrial species. They play the same game that we do, just better. They do the same things in that game that we do, just better.

So yes, grandmasters will be better calculators than amateurs like us (not just depth, also speed and accuracy), but will also have a bigger 'database' of postions and ideas, know more opening theory, more finely-tuned intuition, better understanding of positional themes, less prone to lapses, etc.

and i can say that physicists are not an extraterrestrial species so when i look at the night sky i see the same thing that einstein does.

They probably can only IF it's all forced moves.And 50 forced moves is very very very rare.I can even say it will never happen.Maybe in a composed puzzle position,50 forced moves can arise lol .Otherwise they wont bother with calculating if it's too long.They will just rely on evaluation

Yep, you'll see the same that Einstein saw. But he understood it better.

that's what i meant.

http://www.maa.org/mathland/mathland1.html

http://www.chess.com/forum/view/general/how-many-moves-can-you-calculate-ahead

http://news.stanford.edu/news/1999/april28/kasparov-428.html

Carlsen:http://www.time.com/time/world/article/0,8599,1948809,00.html

About the complexity of some positions:http://wright.chebucto.net/AI.html

There is a video on Youtube where Kasparov says he can see 15 moves ahead in certain positions, I just don't have the time to search for it, but I know it's out there.