Will computers ever solve chess?

Concerning the accusation you make above, I suggest that anyone who cares to to go to the thread mentioned and get my reply to your request to say no more.

They will find that I immediately complied, and left you to go on with the discussion.

Surely, that was a gesture sufficient to trouble you no more.

Do spammers do that?

Of course not.

You have painted me in the worst possible light, but I am not going to reply in kind. 

As for the rest of your reply, which is about the topic, you will have to pardon me, but I think you are saying exactly what I was pointing out.

Engines do sometimes mess up in specific anti-computer positions, but for the most part they just give incorrect evaluations, which is no problem. An engine will tend to play the best drawing moves in a position by will give a positive evaluation. This is due to a problem with the way it evaluates, but it can't be considered a fault unless it actually loses in a drawn position. This is simply because it's not easy right now for computers to understand the 50 move rule. However, before saying a computer doesn't understand a position, make sure that it actually plays losing moves and doesn't just play the correct drawing moves while insisting it will win due to not understanding the 50 move rule.

Also, I've seen some talk on here that computer's are materialistic. Humans are materialistic, not computers. We need to simplify the game into things like point-based material count so we can understand. Computers just look for checkmate and when that fails look for things like activity, king safety, and material as secondary goals. To think that computers simply add up the material at the end of their lines and compare shows huge ignorance about the way chess computers work. Part of what makes computers so good is their extremely accurate intuitions (or rules) concerning king safety and how much material can be sacrificed for an unsafe king. When a computer shows that a sacrifice is +5 that doesn't mean that after the sacrifice white wins back his material plus a rook, it means that the king safety problems for black are a winning advantage according to the computer.

Computers don't have feelings, but frankly I think it's insulting to the programmers of computers and the GMs who assist with the algorithms to think that computers only care about material. Again, it is in fact humans who usually care far too much about material and point value on the board, we should learn from computers about all of the other important elements in a position.

"Also, I've seen some talk on here that computer's are materialistic. Humans are materialistic, not computers. We need to simplify the game into things like point-based material count so we can understand. Computers just look for checkmate and when that fails look for things like activity, king safety, and material as secondary goals. To think that computers simply add up the material at the end of their lines and compare shows huge ignorance about the way chess computers work. Part of what makes computers so good is their extremely accurate intuitions (or rules) concerning king safety and how much material can be sacrificed for an unsafe king. When a computer shows that a sacrifice is +5 that doesn't mean that after the sacrifice white wins back his material plus a rook, it means that the king safety problems for black are a winning advantage according to the computer." xman

What ???

This was just proven to be FALSE. Engines base their next move based solely on it's material evaluation given a numerical value. The higher the number is the move selected. It is humans who are capable of making best moves based on other considerations besides material. Where xman got this notion is beyond me.

+/- numerical evaluations are expressed in a material advantage and have NOTHING to do with King safety.

Thank you for your efforts. Even if we had a thousand move limit--it would still be a draw.  [for one thing we would eventually have to have the draw by claiming a 3 fold repeat of the position]

So not to change the subject but quantum computers seem like the ideal instrument to "solve chess"  Certainly they could expand the table base well beyond the current 7 pieces...

People outgrew the childish need to villify people for being smart?  No new term.  I guess "geek" would be the closest, but that does not have the negative connotations of nerd, so...you might be out of luck.

@xman720

You know what you are talking about.

We have all seen chess engines at work, the score fluctuates, positive for white, zero is balanced, negative for black, when the computer finds a high score it cuts off the search and that's the move.

For the people that do not know that score is the sum of many weighted functions that are run, as he mentioned actual material on the board and then many many others like king safety, passed pawns, doubled pawns, central knights vs corner or edge knights, etc, etc.

It is important to note however that actual material on the board carries most weight in most engines and that is why the engines fail in that example given, they rush to promote, they queen, material balance changes drastically then they cut off the search at the starting position of the example because it is assumed to be winning (the high score caused a cut off)

In chess we use Q = 9, R = 5, B = 3, N = 3, P = 1

In engines they mainly use: Q = 900, R = 500, B = 325, N = 300, P = 100

Called centipawns these starting values for every position are adjusted by positional functions.

For example, bishop pair +25, knight on central square + 20, passed pawn say + 20, passed pawn on say 6th rank + say 40, good king safety (has pawn shelter) say + 50, bad king safety say -50

As you can see all these positional scores add up to a net amount that is added to the actual marterial values but it is usually a fraction of the total value, say 20% to 40% of the final total.

The scores you see with chess engines is one figure made up of lots of different weighted things but usually about 75% or more of the score is for the actual material on the board.

In the example given it is not the starting position that is the problem.

The problem occurred long before that, the engine, say 10 moves back (20 plies back) calculated up to P=Q, then a few more moves just to be sure then cut off the search.

Once the engine gets to the actual starting position in the example it is too late, damage already done.

How do you solve this problem? Well it's not easy.

With so many positions to check engines want to cut off as quickly as possible, but in this case when a cut off is found another deep search, say 20 plies is needed to confirm the win.

But if you have to do it for this position then you have to do it for ALL positions because computers are DUMB.

Catch 22

"The scores you see with chess engines is one figure made up of lots of different weighted things but usually about 75% or more of the score is for the actual material on the board."

 Which is the point. Engines evaluate the material and place a value to each piece and pawn. Some pieces may be given more/less value than it's original depending on it's future potential. But in the final "judgement" the move chosen is solely based on the highest numerical value. For example, in an exchange sac a minor piece will be given a higher value, while the opposing Rook with "no activity, no prospects" is assigned a lesser value. Pieces/pawns material value can change. The engine adds everything up. King Safety ??? If the engine does not see a forced mate in it's horizon, you're trying to say is weighted? That a number is given to "the King is in danger"when it sees avoiding mate is possible and gaining material along the way? I think not. 

"So, once hard drive design comes to reflect the benefits of winding data into spheres, which could be exponential in nature, we might find other ways of using this design.  Spherical houses ... spherical cars ... spherical cities, etc." s23bog

Spherical funny farms !

If the engine has not reached its max depth or the leaf nodes of the current iterative deepening cycle then the engine may sacrifice everything if it calculates forced checkmate, material is not even considered at this stage.

I am talking about the evaluation functions that run on the leaf positions in the game tree. King safety is important in engines, you even get a bonus if you have castled.

You even get a bonus in some engines if you both castle short and your rook is on F1, because it points at the black king.

I have the source code for about a dozen engines on my machine, including Stockfish.

Remember the point of the game is checkmate and in most positions material is balanced or maybe one pawn difference.

That's why these positional functions are important.

King safety is always evaluated at the leaf nodes even when you are way ahead in material.

Forget hard drives: flash memory is already faster, more dense, lower power. The final barrier is price, which is of less relevance to this discussion (except where it is stupendous).

False. In fact this is a difference of engines. An engine is happy with a tactic that has 100 different branches, all of which are necessary to justify it. Indeed, every single move a computer plays involves looking at a tree including millions of positions. Humans cannot be so thorough.

It is true that forcing moves are more likely to be good ones on average, but this is an empirical, statistical fact about chess, not a fact about how humans or computers play it.

I am not sure how you could possibly make such a statement when computers obviously play "promising, good looking lines" all the time, but an example is that sometimes these days computers make sacrifices even when they don't get the material back in the entire tree that they search (they are sometimes able to identify a positional advantage so large it is worth a piece).

"I haven't seen one of those.  I do recall a movie about a guy that got stuck in a Turkish asylum, though.  The masses inside ... the permanent residents ... walked around a pool of water and mumbled to themselves all day.  The American who got stuck there wound up walking in the opposite direction.

It helped him to gather his thoughts, I think."

One Flew Over the Cuckoos Nest was an AWESOME movie   TYVM for reminding me about that gem of a movie!

OK, i gave a position White to play and find the draw which the best chess engines could not find. [but i found the solution]

Now here is a position White to play and win. The best chess engines cannot seem to find the solution but i found the solution just by looking at the chess board.

i am on v2 so cannot do a diagram--so could somebody with v3 diagram  this for me?--it is White to play and find the best continuation.

White K on b3 N on a4  Q on b2

Pawns on a3 b5 c4 d5 e4 f3 g2 h3

-------------------------------------------------------

Black K on b7  B on g3  Q on b6

Pawns on a5 a7 c5 d6 e5 f4 g5 h5

Genyogi you are violating chess.com rules. i do not know if the other guy is violating chess.com rules but, for sure, you are!

In the meantime, could someone with a v3 version post the diaram per post #2730?