Will computers ever solve chess?
Sort:
If you differ with the ponz--U R Wrong!!!
[just kidding]
Jul 27, 2017
0
#1825
Stockfish gives +7 after 35 moves but it's a draw so it appears.
after 1. f3 Kg5 White draws with 2. Be3ch
Why did a 74 year old man figure out a chess position which the best chess engines could not figure out? Because a human can think outside the box and a computer can just follow its programming.
This position was given by a grandmaster and I solved it quickly.
Before that this same grandmaster gave another position White to win--which the best chess engines gave the wrong solution. That one took me about 7 or 8 minutes to find the right line.
So my point is--in trying to analyze chess--a computer [or computers] may look at trillions of positions but there may be some positions which they do not analyze correctly. It is hard to get around this. Maybe the super computers which have speeds of a million times our computers--could have found the right solutions for both chess problems?
Jul 27, 2017
0
#1828
yep. But Black is slowly making an inroad with King in your line. Stockfish gives 35 moves in the analysis shown in the position in bogs post slowly increasing up to +7. (you can click on analysis there). Positive it's been analyzed as draw?
Jul 27, 2017
0
#1829
Problem is comps saw your line and dismiss it. It makes it's evaluation based on material advantage. The eval kept going up and never considered not being forced to stop the pawn.
ProfessorPownall wrote:
yep. But Black is slowly making an inroad with King in your line. Stockfish gives 35 moves in the analysis shown in the position ib bogs post slowly increasing up to +7. Positive it's been analyzed as draw?
i have analyzed it to a draw. Black cannot win if White plays correctly.
However if you show the position and my solution to the average IM or the average GM he will probably agree with me.
[by the way, the grandmaster who published both positions here on chess.com--agreed I had found the solutions to both problems] [and i was the only one to find either solution]
ProfessorPownall wrote:
Problem is comps saw your line and dismiss it. It makes it's evaluation based on material advantage. The eval kept going up and never considered not being forced to stop the pawn.
i can give many positions where one side has a very big material advantage but cannot win. This is a problem these chess engines have--there is more to chess than a material advantage.
Jul 27, 2017
0
#1832
Stockfish gives 3 lines all with the exact same evaluation. It's stuck in a "Time warp" moving the Queen all about and sometimes the King. Just can't figure it out. I would think this "issue" could be addressed by some programmer. Certainly engines can never begin to "Hope" to solve chess with this glaring deficiency.
When chess engines become self-aware--they should be able to solve such problems?
camter wrote:
Some of the crazies here are nerds who say, basically, that the problem boils down to too many positions, and not enough time or computing power to solve them in 1000 billion years,
If so, I agree thgat no "solution" is available.
I have a Rook and King at one end of the board, and my opponent has a lone King at the other end.
In realistic terms, I know I have a win for certain.
Would they agree or not that the problem for such an instance is solved?
Or, is it not solved because I cannot give the "analysis" required without giving every possible move and answer? Is that analysis required for that case to show I have solved the problem for the general case od K+R vs K?
Oh, no, do not be ridiculous, @camter, we have an algorithm!
I agree and I know the algorithm, because the technique is simple.
But, what if the positon we are confronted with has 10 chess pieces a side, and has consequently a very large complexity?
As long as I can "describe" a technique to the computer for that position, I can get a solution.
That is why some nerds are trying to use relational database ideas, and stuff like that, to solve Chess. Good luck to them, and it may work in lots of cases.
But, I cannot see why that is necessary or even possibly desirable.
Surely, brute force, capacity and time are all that is required.
There are gazillions of possibilities in a game of Chess, and because of that, there is a thing called the horizon effect, because of the limitation in time and resources. So, we have the eval which is insufficient for the time available thereof.
That eval is a tool, or number, which helps with a concrete solution, but a help only, as the present computer cannot see beyond the horizon, because somewhere beyond that horizon lies a mate for Black or White, or it does not and so we have an impasse, which we can call a draw, if we allow (by arbitrary rules in most cases) that a position has insufficient material, a repetition of moves, stalemate, or a 50 (pick your number) move rule, or "agreement".
You realize that people stopped saying "nerds" in the 80s and 90s, right?
Jul 28, 2017
0
#1835
camter wrote:
Oh, no, do not be ridiculous, @camter, we have an algorithm!
I agree and I know the algorithm, because the technique is simple.
But, what if the positon we are confronted with has 10 chess pieces a side, and has consequently a very large complexity?
As long as I can "describe" a technique to the computer for that position, I can get a solution.
That is why some nerds are trying to use relational database ideas, and stuff like that, to solve Chess. Good luck to them, and it may work in lots of cases.
But, I cannot see why that is necessary or even possibly desirable.
You really have no idea how chess engines work do you?
See sections above in red, that is how they first tried it with chess back in the early days during the 60's and 70's last century.
It sounded nice in theory but did not work in practice, too much complexity, too much logic, too much code.
Now they just use brute force and if you use various techniques you can prune the game tree a lot which means you can go to much greater depths and the engine will solve the problem indirectly.
Then they throw in 2 other tricks:
Opening books so the engine feels like it plays chess and avoids having to write LOTS of code (like you want to do as per above), and
Endgame tablebases so the engine can peek over the game horizon and once again avoids having to write LOTS of code (like you want to do as per above)
Get your facts right, or at least know what you are talking about, before you post and criticise people.
I played this out with three different engines Stockfish, Shredder and Komodo. It's a draw if both black and white make the best moves. I took it to the 50 move limit with every engine!
Jul 28, 2017
0
#1837
btickler wrote:
camter wrote:
You realize that people stopped saying "nerds" in the 80s and 90s, right?
Really?
What is the current term if you would be so kind?
Can't wait until the millennials are old, and their progeny pokes fun at them. The time is coming soon enough. Woodstock seems like yesterday.
Jul 28, 2017
0
#1839
game_designer wrote:
camter wrote:
You really have no idea how chess engines work do you?
See sections above in red, that is how they first tried it with chess back in the early days during the 60's and 70's last century.
It sounded nice in theory but did not work in practice, too much complexity, too much logic, too much code.
Now they just use brute force and if you use various techniques you can prune the game tree a lot which means you can go to much greater depths and the engine will solve the problem indirectly.
Then they throw in 2 other tricks:
Opening books so the engine feels like it plays chess and avoids having to write LOTS of code (like you want to do as per above), and
Endgame tablebases so the engine can peek over the game horizon and once again avoids having to write LOTS of code (like you want to do as per above)
Get your facts right, or at least know what you are talking about, before you post and criticise people.
We are talking abot solving the game, are we not?
So, if I know very little about the actual working of a Chess program, how does that affect my argument?
And might I say that your reply seems to be lacking a little in logic, in spite of you seeming an expert on subject of chess solving programs.
Please tell me if such a thing as an "horizon" exists, because if it does not, why do engines give evaluations, and not claim Mate in 10,20.30 or 250 for whichever side, or, in any other case, claim in some form that the game is destined for a Draw with best play?
Jul 28, 2017
0
#1840
camter wrote:
We are talking about solving the game, are we not?
So, if I know very little about the actual working of a Chess program, how does that affect my argument?
And might I say that your reply seems to be lacking a little in logic, in spite of you seeming an expert on subject of chess solving programs.
Please tell me if such a thing as an "horizon" exists, because if it does not, why do engines give evaluations, and not claim Mate in 10,20.30 or 250 for whichever side, or, in any other case, claim in some form that the game is destined for a Draw with best play?
First things first, I am not an "expert", you only see "experts" after the fact, never before.
Next thing, I was on another persons topic about relational databases a few months ago and you kept interrupting with spam and useless comments, I asked you to stop it, and here you said "nerds, blah, blah, relational databases, blah, blah..."
Last thing, in chess the game horizon is the current max depth that the engine can see, it varies with the setup and engine.
If it is currently 30 plys (15 white moves + 15 black moves) for a particular engine then the engine is clueless about ply 30 + 1, it's game horizon is 30 plys.
It may evaluate a position at ply 30 as winning for white after white just captured the black queen but will miss checkmate by black the next ply on ply 31.
Ok, here is the solution. [please diagram it for me]
White moves first
1. f3 a4 [this is the main line]
2. Kf2 a3
3. Kg3 a2
4. Kxh2 a1=Q
5. Kxg2 and the position is a draw.