How good do chess engines play chess?

@wafflemaster: our posts keep crossing each other. I understand that a program is a result of the way programmers build their evaluation program. Normally I would say capacity depends on speed of information processing. The best GMs have not a much bigger information processing capacity per se then other people. If settings would be possible for a chess engine to have an equivalent information processing capacity, then would chess engines and human play become comparable.

Like I said before, if you make the computer calculate as slowly as a human, you should also take away the human's ability to think about chess positions.  Then yes, I think you could compare them.

The problem with this though is that both players would play nonsense moves.

It's not very useful, but I can't think of a better way to make it so you can compare humans and computers.

@wafflemaster: I agree, because the ways to find moves are incomparable. Thank you for your time.

Loek  You want to know how lower rated humans can help chess engines such as in Centaur Ches?  I am now experimenting with Centaur Chess per the Ponziani Challenge.  How have I helped my new stockfish engine?

1. I purposely made it think or analyze about some particular postions and lines which had not been seen before. For example there was a recent game FirebrandX vs Brian Wall in the Fraser Variation of the Ponziani.

Brian Wall is a leading experent on the Fraser Variation and FirebrandX is the current US Correspondence Champion.  The game ended in a draw but even before I received my free stock fish I had ideas of improvement for Black.  These I force fed into stockfish and my ideas worked fairly well.

[so far with Black pieces I have 4 wins and a draw in this Centaur Chess exhibition]  This would not have happened if I had not forced stockfish to analyze seeveral of my lines [suggested improvements]

Also, as I was playing other players, I noticed that some lines by stockfish just did not work and thus I directed Stockfish away from those lines and gave it one or two additional lines. Thus with my human help there are new lines created which stockfish, alone, had not created in the past.

Another thing humans can do to help stockfish [or any chess engine] is to program opening lines into it.  also some centaurs add whole data bases of thousands of games played in the past.

Computers are artificially intelligent. They have an extremely stupid way of evaluating a position and choosing candidate moves, but by merely using brute force calculation they seem to come up with an intelligent solution.

Humans use real intelligence to evaluate a position and choosing candidate moves, but lack in calculating power.

This is the reason why computers play better in complicated positions and humans play better in quiet positions.

A setting to compare humans with computers might be to give them positions to solve by which a certain amount of the positions have a tactical solution and the other amount of the positions have a strategical solution.

@ponz111: I know Brian Wall and have seen the game against FirebrandX. I have asked attention for his remarks on the Fraser variation in the thread about the Ponziani, don't you remember?

Interesting remarks how you have helped Stockfish.

@ultimateichigo: computers can not work using associational networks like humans do, but that does not make it not intelligent by definition. Different, that is. Just as computers can not explain how they come to their conclusions, we can't.

There are computer programs that can adapt their own behaviour based on processing new experiences. That is equal to learning. I consider learning from output a necessity to define behaviour intelligent. If a computer program is capable of reacting to new situations (a chess engine is) and can improve its own behaviour based on past experiences (they can), then is it imo real intelligent behaviour. It is quite challenging to accept other forms of processing as other roads to intelligence, don't you think so?

Loek  I have been giving Brian Wall a lot of attention on the Ponziani thread and his Fraser Line.  I have found improvements on the game published plus other improvements.

@ponz111: I will check it out, because I use the Ponziani as my answer to the Petrov.

You forgot Schumacher! 

However, as already said before, Man vs. Machine is a win for machine. But I guess that Houdini would (almost) never win any tournament, like Tata Steel Chess, because he'd never win and never lose.

And yes, human+engine is surely better than everyone else. Check Ivanov-stuff. 

@Petrosianpupil: No engine expert here but I've heard good things about Stockfish being more "human" in its' play.  I've been happy enough with Houdini 1.5 that I haven't bothered with anything else.  If you're serious about finding a better engine I know that there have been threads in the past that seriously debated the merits of about a dozen mostly free engines.  You could do a forum search and there are free downloads on site as well.  Maybe load up 5 or 6 in Fritz and give them each a game you know to analyze and see what you like best.

Y

Computers play chess with brute calculation, calculating for example, 70 moves into every line, this means that for every move possible in a position, the computer calculates so far ahead it's humanly impossible to match it. This said, a computer does not have candidate moves- just brute calculation until it finds a move which results in an analyzed position where it is better. Computer strength is simply improved by improving the overall speed of the computer's calculation ability which enables it to go deeper into more lines, increasing the likelihood it will find the best move in a position.

If being able to play perfectly from any position would mean lifting a whale, then humans would be like mice, engines like cats, and centaurs like dogs. While each is stronger than the former, the strength of any of them is meaningless when faced with the task to lift the whale.

Fortunately more than 99% of all positions are clear wins for one side or the other without having to play perfectly.

Centaur Chess players at the highest level can sometimes play a perfect game of chess. So can grandmasters. It has been done literally thousands of times.

What was his thesis and were his conclusions about AI and ant colonies? Can you tell it in a few words?

I have Sjeng, although I hardly use it. It is very tactical. The only time I have won from it, was when I played a closed position in which I had sacrificed a rook for a knight. My opponent made some bad moves and lost very quickly. Then I decided to play the situation against Sjeng. He played worse than my opponent. I also had once a won end game in which I had three very advanced pawns against two pieces. I could calculate the win to the end, he could not see it.

First I wrote Michael Schumacher, then I decided to choose Vettel, because he is doing very well the last years. When I replaced Schumacher with Vettel I already thought about you, but hey, Vettel is the new kid on the block.

Of course is a chess engine performing better than a human. No doubt about that. I was wondering if it was possible to have the chess engine play with a kind of configuration that more or less equals the strength of a human mind in order to compare the strength of a chess engine program with chess players of different levels. Some standard RAM/Cache, ply depth etc..

Wafflemaster convinced me that it is not possible. The remark of Chessman_47 that you can win from a program when you are smarter than the programmer can not be taken serious. Brute force calculation adds more power to a chess engine than the intelligence of the programmer.

Your reasoning is correct. It is only not known if for perfect play the strength of lifting a whale is required. That is a big if in your equation. I would say that the level required for perfect play is at its most undetermined. :-)

No, seriously, I think that a the level of chess is already quite good. There are a lot of positions that are too complex, but there are also very much positions that can already be understood with the current level of knowledge.

I think that the end position of this game is a draw by all means:

(I do not pretend that the end position is reached after perfect play from both sides.)

What do you think? Is this a draw or not? We have already the 6 pieces tablebase and have therefor definite knowledge about a lot of situations.

In the you + Carlsen combo, you can't help Carlsen.

Sure you can...you can suggest a move, and Carlsen will say "that's a patzer move" and play something better.  Viola, you have helped Carlsen elmininate a move from consideration ;)...

Haha, that is exactly the essence of my analogy. And if I would say by accident the best move, then will I probably have the wrong reasons for it. :-)

When chess engines are developed, Grandmaster assistance is given to the programmer. Especially super chess computers like deep thought, deep blue, or rybka