| From | Message |
|---|
hortstu
27-Sep-09, 20:27
|
 Is there such a thing as a
I understand that it isn't really perfect, even by my own definition, but I want to know what others think...
Assuming 2 super chess computers are playing each other, w/o time constraints, and they make the statistical "best move" every time. Technically they are playing into each others hands the whole way. I assume the reason that chess masters still beat computers is because they don't always make the best move and this makes it harder for a computer to predict the outcome of every possible move at every possible turn.
Does such a game exist?
If so where can I get the notation or annotation on it?
Have the masters ever played it?
Is such a game common chess knowledge and maybe I don't know about it?
Would such a game always end in a draw?
Thanks
|
|
Such a game would mean that the game of chess has been solved, fortunately this has not happened.
|
|
The problem for computers is that each new move they have to compute in a sequence takes increasingly more time to compute. This is already limited to a point by having opening books which list the most commonly played openings. The earlier you want the program to be on its own, the more it has to calculate in order to ascertain that a line is better than others. If you want for example for a program to decide between 1. e4 and 1. d4 it likely means that you'll have to let it compute ~30 move deep lines, meaning 60 ply lines. That is currently impossible with the available technologies. So what we have now is programs that will chose among several openings from their opening books, and have a built-in decision point at which they switch from playing book lines to calculating on their own. And those calculations stop after a certain point because the next steps would take too much time to be of any help. If going on one more ply means having to take a few days, one more ply more than a week, one more two weeks, you can see that "no time constraints" is not a realistic way to play chess.
This is all greatly simplified, but I hope you get the point. It would take too long to "be sure" of a move, and even then there's nothing guaranteeing you that you're actually right.
|
|
As for how people can beat computers even without all that computing power, it's because a human can say: f7 is weak, so my plan is to attack f7; I have to activate this, this and this piece for the attack. The plan might take 20 moves, 30 moves, 40 moves, what with having to counter the opponent's moves too. But the human will not analyze Kb1 (sometimes wrongfully) when the action is on the kingside. The computer cannot say "I only analyze this and this line". It might decide early on that Kb1 leads to a better position for the other part, but on the other hand it might take a loooot of time to see this. When it does, the time spent on this is already lost.
Of course, computers become more and more powerful, and they rarely make any tactical errors, while humans are more prone to blunders. Even if you play at your best in most games and you share points with the computer in those, nothing counters human blunders, the computer will not give you any "free" points.
|
|
It could be a while yet before chess is 'solved'. Check the last couple of paragraphs of this
article.
news.bbc.co.uk
garos.
|
cascadejames
28-Sep-09, 22:37
|
When a computer is taken out of book early
On this topic, Check this interesting page and especially the link to
"Defending Humanity's Honor" The continuing story of how computers can't play chess
www.xs4all.nl
Thanks to obsteve for originally finding and posting this interesting link over in the philosophy,
poetry, and art club forum
gameknot.com
|
hortstu
29-Sep-09, 08:13
|
OK great responses and I appreciate the interesting links as well. The link garos posted and the first one by cj made it clear that chess is by no means solved, and not even close to it at this point in time but that isn't exactly what I'm thinking of when I say "perfect game."
Maybe a perfect game is just one where neither side makes a mistake or blunder throughout an entire game? In baseball a perfect game is one where a team doesn't allow any walks, hits, errors, or hit batters... basically no baserunners... I hope the non baseball fans will forgive this analogy... What actually occurs during the game is different every time but to be called perfect it has to have the same end result and not just a win. (there are no draws in baseball.)
So I guess I'm looking for a definition to a perfect game as much as I am an example of one. At this point in time it seems all the previous responders would say one exists but no one knows what it is yet, and we won't until all the billions and billions and billions of positions are thoroughly analyzed by some super computer yet to be invented.
|
|
Yeah, since the number of moves and positions is finite (if you include the 50 move-rule) we can
say that there is a "perfect game". The problem is we have no idea what it is, or whether white
wins it, or it is a draw, or even if black wins. The latter might seem unusual, but think of it in
terms of information: white moving first "reveals" his play to black who then can make a more
informed decision.
|
|
What is the definition of “Perfect Game”?? A game where both sides do not make mistake? If so, this is only possible if this is true: “Perfect games will always lead to a draw” (This is my own believe but no one can proof it).
If “perfect games will always lead to 1-0”, then it is hard to define the “perfect game” because if BL has too loose at the end, what is the different of losing in 20 moves or losing in 60 moves? Will the longest defeat be the correct move for BL? Or BL doesn’t have any correct move at all?
Now let’s assume “Perfect games will always lead to a draw”. What if you have the options to force a draw in 40 moves or to do it in 43 moves (from one side perspective), which one is the “perfect moves”? Can both be considered “perfect moves”? If so, then perfect move or perfect games are not a mystery at all…
Imagine why higher ranked GMs often have to play weak defenses such as Ben-Oni, a defense that IMO will lead to defeat if played by Gods. In GM level, finding “perfect moves” in a simple position is often easy enough for both sides. So to avoid draw they may decide to create complex position (or unknown position) by making “imperfect move” (which is a blunder from God’s perspective).
So, playing chess is not about finding a perfect game. When 2 superGMs do not want to take risk and be okay with a draw, they will choose simple drawish lines. And usually both sides will notice this intention so instead of wasting time playing the “perfect game”, they will just shake hands.
And don’t worry about the future of chess. Even if the computer will be 1000x faster, it doesn’t affect if the time control is shorter. In the future people will have no time playing 40-moves-per-2-hour chess. FIDE will shorten the time control.
But indeed there are so many “secrets” to chess understanding. That’s why we still often hear questions like “what is the best defense?” from beginners. So, I will explain my reasoning in choosing my opening against 1.e4 to show how it relates to the so called “perfect game" (I considered the Sicilian, the Caro-Kan and the French for my opening study).
I believe that most players can be considered either positional defending players or tactical attacking players. I believe I fall into the first category. People often amazed to see how I can survive defending difficult positions, and at the same time I often resign a winning attacking position because I thought I have failed all the sacrifice, and couldn’t see the forcing mate. So, the first criterion for my opening is whether it suits my character. The French is not but the Caro-Kan is.
Next criterion is the theoretical strength of the opening. Many GMs avoid what so-called “drawish” openings so they are not popular. But in my level drawish openings are not drawish at all, and often the drawishness is the sign of “perfect opening”. I value these lines highly. The Caro-Kan and Sicilian is, the French is not.
Next criterion is the learning experience. The Sicilian is a perfect opening but it requires master level skill to play it well. The end positions are often too difficult for players below 2000 to comprehend. What does this mean? It doesn’t mean that you will lose more often because you’re not the only one facing the difficult position but your opponent too. The problem is when you don’t know too much of the positions, you don’t learn too much, and you improve slowly. The Caro-Kan and the French is good, the Sicilian is not.
So I leaned toward Caro-Kan, but I have played the Sicilian for years since I had first access to opening theory. So without using logic, I stick with the Sicilian! I just have to choose the variation!
Fortunately, the best candidates are in …d6 variations (Dragon, Scheveningen, Classical, Najdorf). The Najdorf is theoretically the strongest but the most drawish in the Sicilian. The Scheveningen is theoretically inferior but it ensures genuine complexity. The Dragon is the most inferior but it suits certain character (those who is good and like to have the initiative to attack even if they have to sacrifice materials), and it will give the highest learning experience.
You see, these 3 variations are “Perfect”! (None is perfect for my character, so I chose the theoretically strongest, the Najdorf, and I choose lines that are more suitable with my character and playing technique).
|
|
Chess solved <> perfect game
Chess is only solved, when for every move a reply exist that leads to win. A winning combination starting from move 1. Or like you will, an opening book till the end.
This is quite different then the "perfect game" that two supercomputers without time limits can play.
|
|
When chess is "solved" most experts think that it will shown to be a draw. It is also possible (although not very probable) that the initial position is a zugzwang and every possible move that white could make leads to a loss.
|
|
Please log in to your account to play chess online!
