Ive always been intrigued by chess, but Im very bad at it. However, ever since I created my own chessvariant (free download here: http://www.chess.com/download/view/economy-chess spam spam) Ive gotten more and more interested in the game and watched quite some videos and 'clickthrough games'.
The most interesting to me are the engines. I just like that they do the weirdest counterintuitive moves that pay off many turns later. Also I am fascinated by the perfect game, if that exists.
So Ive been looking for a game that comes closest to that perfect game. In other words: Ive been searching for a game with the strongest engine, that plays a game against himself with a very long time control and calculated on a supercomputer.
With not so much succes...
Do you know which game comes closest to what I described? A link to a video or a 'clickthrough game' would be awesome. Thanks in advance!
The more you know about how computers play chess, the more you'll realize why they're just as far away from a perfect game as a people are (relatively speaking).
Perfect games happen for 1 player (not the other ;) when one player is much better and so all his moves are demonstrably the best.
The strongest engines are available for free download (I think they reserve the latest model for pay though :p And you can have them play matches against each other.
It's fun to watch (I've done it myself) but completely indistinguishable from watching two human players who are much better than you. Just as many mysterious and amazing moves.
A "perfect game" is a fairly meaningless concept anyway. It's not like getting a 300 in bowling...in chess it's a lot more nebulous.
I do believe chess can be solved, resulting in a draw or a forced win right from move 1. We're miles away from that. But I'm curious to what the current status is.
In 2011 Junior won the WCCC, so I guess that is the best engine. Have they let him compete himself? With a timecontrol of, say, 2 days/turn on an IBM supercomputer?
Or perhaps another engine but under similar conditions. Ive searched a bit on youtube and found some videos but I dont know what timecontrols where used. And I dont really know how to search through the games of chess.com and chessgames.com.. there has got to be a game like I described out there but I havent found it. Hope others can help :)
Well, presumably it would not be "a" draw or forced win from move 1, but a number of them (depending upon the initial moves). In fact, it would probably be a vast number of them. And bear in mind that your belief is a belief...and only that.
Here's a ranking of top programs: http://computerchess.org.uk/ccrl/4040/rating_list_all.html
Chess is solvable, but there are some serious roadblocks. Say one chess position and the evaluation for it took 1 bit of computer memory to store, and you had a storage device capable of storing 1 bit of data per atom. Even with a such a storage device a large as a planet you can't store all the positions. In fact you're still many orders of magnitude away.
Things like this make a proper solution impossible in practical terms.
So are you sure it's solvable?
I'm always a bit leery of "____ is a big machine" claims anyway, since I seem to recall that something called quantum theory came along and loused that picture up a bit.
It hurts me when you say a chess position can be represented with a single bit, even though I know you probably know it isn't true and are just saying it to demonstrate a point.
Gee browni, I think you're taking this thing a little hard...
Ok, what I mean is there's nothing in the laws of physics that makes a complete solution impossible.
But there are tremendous practical difficulties.
Sure a quantum computer could work though a solution, but it seems to me that it would be difficult tracking what positions you've explored. Maybe there's a way to provide a full solution without storing all the analysis, but my initial impression is you'd need this to be able to verify (and guide) your computation.
And sure there may be shortcuts to the 10^120 positions (or whatever it turns out to be). But even in fractions that's a huge number.
And yes I realize that's impossible browni, I'm just trying to get the scale across. I suppose I could have said more information than there is matter in the universe, but that's so big... (and I'm not sure it's true exactly :p)
true, havent thought about the possibility of multiple solutions. I wrote 'believe' intentionally though. But the faith is strong :P
@Wafflemaster: I see, but isnt it possible to cut away humongous chunks of the even humongouser amount of possibilities with smart programming? Perhaps just enough to get it solved before humans extinct..
@Browni: Im sure that was for making a point.
And there are after all already a large number of demonstrable repetition draws coming right out of the opening (and sometimes even still in it). Those will have to serve as a part of the morass of background footnotes behind this alleged "perfect game" that someday the computers will be able to pull off.
I guess the first step is to map out the legal positions and once we have that we can estimate the kinds of reductions available by trimming certain branches.
But we don't even have a good guess for the number of possible legal positions yet... so I wouldn't hold my breath :)
thanks for the computerchess.org link!! that will get me as close as possible to what Im looking for.
Will check it out after my night shift while sucking some cloud out of the bong :P
Ok I take it back, hold your breath.
Oh don't do that, you'll end up looking like this:
(This was the episode when, according to Barry Williams' memoir, Greg Brady appeared stoned on camera.)
I know what you were trying to say, but the dog has to chase the squirrel if you know what I mean.
The most interesting to me are the engines... Also I am fascinated by the perfect game, if that exists.
Download Houdini, the strongest engine on the planet, and have it play itself at a very, very slow time control. That's as close to the perfect game that you will find.
Take a look at this nice video of a game between Houdini 1.5a and Rybka 4, played at the TCEC tournament (kind of unofficial world championship) early 2011:
This could be the best engine game ever. Houdini sacrifices one - two - three pawns in a queen-less middle game and then plays the quiet move 23...Kf7!! - having full positional compensation for 3 pawns.
Somebody wrote the following comment to the video: "The stronger the chess engines become, the more random their games will look to us." :)
Thanks RHoudini! I will take a look at it.
I tried the computerchess.org link and downloaded the games from their strongest computer. However, when I tried to open the .pgn file in Chessanalyse 2.6 it gave an error: Ambigious move :(