Computer Analysis is CR*P!!!

it means a computer able to calculate 50 moves ahead

I think you are right in your statement and don't think that it is ever achieved. Is it? 50 moves, or 100 half moves, would mean that a computer must evaluate an estimated 255+E152 positions.(*) Thats 255 with 152 zero's behind that.  

(*) calculated with an average of 35 possibilities for every half-move

Yes... and no.

There is no way to avoid having to calculate a number of moves exponential with calculus depth, but there are ways to reduce considerably the calculation tree to a smaller exponential. A simple example is transposition tables. Imagine you calculated what is White's best move after 1.d4 d5 2.Nf6 Nf3. Then, you already know what to play in the event of 1.d4 Nf6 2.Nf3 d5 for instance.

A more sophisticated, and way more efficient example, is the alpha beta pruning, coupled with the search of so-called "killer moves". To simplify, say that you have seen that move A is refuted by move B in some variation. Then, when you calculate move A in another line, you try move B first because you assume it is more likely to refute move A than other moves ; and if it does refute move A, you do not need to calculate the others.

Andy, if I was friends with Geena Davis, I certainly would NOT introduce her to you. I would not introduce any friend of mine to the likes of you. If I did, I would surely lose friends. As for a duel, I think you are already losing the battle of wits.

Wow, Irontiger and Funwithchess, you two are quite the computer programmer and mathematician experts! I never knew this thread would eventually bring in such an audience, especially after the first insultingly simple minded comment! I thought this would be a dead thread. Thank you for enlivening things!

And thank you, too, Tetsuo Shima! That was clever of you to see the computer play itself. Apparently in some positions, it knows it has a won game but just doesn't have a plan to win it. Clueless, or as Irontiger said, non-existant.

Tetsuo, I agree that computers do not use brute force of calculations, and that a computer needs to be able to evaluate any particular position as winning or losing and by roughly how much. Still, I think the Chess.com computer can know it has the advantage, but not know where it is or how to exploit it, unless there is a tactical line somewhere. Without tactics, this computer is utterly planless and clueless.

No, computers do not have any positional knowledge.

The basic idea is to use a minimax algorithm. You need one way to calculate a rough static evaluation (ie without looking further moves) of the position in the end of the calculus tree. In chess, the obvious way is to set this is to say ev = number of white pawns + 3*numbers of white minor pieces +... - number of black pawns -... +10000 if Black is checkmated -10000 if White is checkmated.

Then, you search in the tree of the possible moves the optimal path for the side you are playing, say White, who tries to make the value of ev as high as possible, while at each move Black tries to make it as small as possible. So the ev value for a node where it is White to play is the same as the highest value of the nodes under this, and the lowest if it is black's turn. When you reach the depth you specified ("calculate XX half-moves") you stop using this and use the static rough evaluation instead.

The only place where you can add "positional thinking" is in the static evaluation function, by such things as +0.3 for rooks on open files, etc. But I wouldn't call that much more than brutal force.

thats what i mean, but its still positional evaluation, nowere near master level but i guess far better then that of most other players. And if the position is not too complex, sometimes it could be very correct i would suppose.

I use the computer analysis quite often in my online games and I have found it does little to really help me.

The problem is that it's just so precise -- even in games where I felt I played pretty well, it finds a ton of mistakes, many of which were not really mistakes, as the moves I played still led to an eventual win. So it says "Oh no, you made a mistake" and proceeds to show me a line that I should have played, leading to a "decisive advantage", while the line I actually played also leads to a "decisive advantage" ... however maybe the line it gives evaluates to a material advantage of 6.76 while my line is 6.01.

So yeah, I should have played the line it gave, but I never, ever would have, because it leads to a crazy tactical position that I would be totally uncomfortable playing. In some cases I don't even understand the reasoning behind the move it gives.

I understand your frustration, but I guess the whole point of the computer is to give you what it thinks is best. I can't give you every possible winning position as it would churn out too much data for each position and probably crash the server

I think the computer should be smart enough to understand choosing a line +15 or so is not a blunder when you can get +17. If you manage to win a won position you shouldn't be bothered looking at what the computer suggest though. Winning a won position in as few moves as possible isn't really a skill you need

depending on rating level,  (of course cheryl's rarting s a lot higher than mine), but down here in my rating neighborhood, you can get away with blunders because sometimes the other guy does not see it and fails to make the winning move. i think the acid test would be to play the computer on chess.com and set the strength to 5 seconds or higher and see how you do. i can beat it at 1 or 2 seconds per move but i think that at cheryl's level of play (i've been crushed by her more than once) 5 seconds would be appropriate. i play the computer on my android phone chess.com app. just tap on "play computer" on the first screen that comes up.

No, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, and no !

The computer is precisely best in middlegames where tactics lurk around, NOT in the early opening. (except if he has a theory book implemented, but those are made by humans, not by computers)

Putting aside that the chess.com free analysis computer is limited in strength so that the only useful purpose of it is blunder-checking (but you can get strong engines for free online) ; computers are already better than GMs at "skill" or "positional chess" or whatever you call it : when you calculate 50 moves ahead, you can see the benefit of open files in terms of material gained.

Where the computer cannot help you human, it's because it does not give reason for the moves (because the reason is "I calculated this to be best"). Not because he is limited in his calculation. An average patzer like me in still miles away from beating any of those dumb machines even once in a while.

LMAO at asterisking out CRAP. 

Jamalov wrote:

i think the acid test would be to play the computer on chess.com and set the strength to 5 seconds or higher and see how you do. i can beat it at 1 or 2 seconds per move but i think that at cheryl's level of play (i've been crushed by her more than once) 5 seconds would be appropriate. i play the computer on my android phone chess.com app. just tap on "play computer" on the first screen that comes up.

Well, Jamalov, I just tried putting your theory to the test. Unfortunately my iPad does not support Java and so I cannot play Chess.com's computer. At least not with what I have.

When computers first came out, I found a simple method to beat them. Take them out of its opening quickly, lock the position, and it will eventually move one piece back and forth. Then I reorganize my pieces to gain further space or the decisive breakthrough. Very easy.

I ended up selling that $200 chess toy for $10.

I can play the computer on my iPhone but it makes immediate and bad moves. I can't figure out how to slow it down and play better. It stumbles into one pawn fork after another, neglects to recapture, and sometimes sacrifices without any reason. I would not even say it plays chess. It just moves pieces.

Please try again today. Good luck.

Irontiger wrote:

LadyMisil wrote:

When computers first came out, I found a simple method to beat them. Take them out of its opening quickly, lock the position, and it will eventually move one piece back and forth. Then I reorganize my pieces to gain further space or the decisive breakthrough. Very easy.

Please try again today. Good luck.

No thank you. I do have respect for advancing technology. Plus I am not about to spend another $200 when I found real humans to play on Chess.com.

I agree that computers have serious weaknesses. THe strongest engines still give many endgames as = that are dead lost. I would never advise anyone to learn about the endgame by following computer assessments, unless it is a clear book position that is known.

That being said, I could outcalculate computers in the 1980's. I cannot now.

I was just wondering why would you analysis a won game using chess.com computer analysis.  Would not analysize a lost or drawn game is much better.  Just wondering.  Maybe you had your reasons.  I usually do analysis/annotation for my the game i lost by myself & send to my friends if i don't understand something.  If it still does not satisfy me, then only i send it to computer.   

Anyway, nice sac.  BTW, i agree with you playing safe but don't agree of you doing an analysis of a won game.  That is just my view which i am entitled to. ;)