And do you think nature or God whatever spent how many years to create us ? In comparison it's not much for IBM 
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Humans vs computers, chess vs trading
- Thread starter Cutten
- Start date
Those of you who argue that in chess there is only a finite number of states, you are correct in principal. HOWEVER, the number of possible states is greater than the number of electrons in the known universe, which is why computers are just as helpless as humans in a brute force approach to "solving" the game. It's not all that trivial. If the number of states is finite, but so big that nobody could even imagine a way to treat them all, what difference does it make?
Quote from Lobster:
Those of you who argue that in chess there is only a finite number of states, you are correct in principal. HOWEVER, the number of possible states is greater than the number of electrons in the known universe, which is why computers are just as helpless as humans in a brute force approach to "solving" the game. It's not all that trivial. If the number of states is finite, but so big that nobody could even imagine a way to treat them all, what difference does it make?
True, the number of possible complete games is staggering - but the number of legal/logical responses to any given distribution of the pieces is limited. For instance, the possible responses to any opening play is only 20 ( 1 of 8 pawns up one/two squares or a knight out ). If we assume that our players are good players and playing to win, we can eliminate all the illogical and/or chaotic games and arrive at a greatly reduced number.
As a chess enthusiast I have no problem accepting the notion that brute force analysis of logical possible combinations made after each of the opponent's moves is doable by a computer.
Obviously, since it took this long to develop a world-class chess playing program there's more to it than that, but the amount of brute force required is probably nothing spectacular. I don't know how it's done, but I should think the computer has a huge catalogue of masters' games to refer to and is familiar with all the tricks.
Quote from harrytrader:
I don't agree as for market see http://www.elitetrader.com/vb/showthread.php?threadid=24752
http://www.europhysicsnews.com/full/13/article5/article5.html
Plectics: The study of simplicity and complexity
Well, I was interested in these complexity science once for a while, and read bunch of books. I don't know why you have quote this study field.
In principle, Chess board state is countable, and yes currently, it's impossible to calculate all, but eventually, it IS possible when the computing power become powerful enough. I don't know if it's quantum computing or other, but when the time come,it will be possible.
Chess is Math, and there is no hidden uncertain factor there. All of the required information is completely dislosed on the chess board to anyone who observes it. Our limitation is just lack of power of computing it.
On the other hand, the market is not Math. You can never process complete set of information there. The only way to deal it mathematically is to process statically with subjectively chosen limited information.
Quote from rodden:
True, the number of possible complete games is staggering - but the number of legal/logical responses to any given distribution of the pieces is limited. For instance, the possible responses to any opening play is only 20 ( 1 of 8 pawns up one/two squares or a knight out ). If we assume that our players are good players and playing to win, we can eliminate all the illogical and/or chaotic games and arrive at a greatly reduced number.
As a chess enthusiast I have no problem accepting the notion that brute force analysis of logical possible combinations made after each of the opponent's moves is doable by a computer.
Obviously, since it took this long to develop a world-class chess playing program there's more to it than that, but the amount of brute force required is probably nothing spectacular. I don't know how it's done, but I should think the computer has a huge catalogue of masters' games to refer to and is familiar with all the tricks.
You don't gain anything by eliminating "obviously stupid" moves. You probably know of games where a queen sacrifice leads to a mate in 5 or something like that. Any computer that analyzes the game only 4 moves deep will capture the queen and be doomed. The reason why there are no examples of a queen sacrifice that results in a mate in 25 is very simple: No human and no computer could ever think 25 moves ahead. Let's say there were only 3 legal moves in any given situation, or you might say there are only 3 REASONABLE moves in every situation. Than means 9 different possibilities for the first move (white has 3 moves, and for each of his moves black has 3 responses). For the first 2 moves we get a total of 9*9=81 possibilities. For the first 12 moves it is 9^12 possibilities, which happens to be a few hundred billion, just about as much as a computer can process in a reasonable amount of time, say a minute. That's why most chess programs don't think ahead more than 12 moves. But for the first 25 moves we have 9^25 possibilites, which would take any computer several billion years to go through.
But the computers are programmed to play book openings, so they don't have to figure any of this out. They don't start "thinking" until the game has left opening theory. And I guarantee the opening book in a word class computer is greater than any humans opening knowledge, even Kasparov.
Quote from Lobster:
You don't gain anything by eliminating "obviously stupid" moves. You probably know of games where a queen sacrifice leads to a mate in 5 or something like that. Any computer that analyzes the game only 4 moves deep will capture the queen and be doomed. The reason why there are no examples of a queen sacrifice that results in a mate in 25 is very simple: No human and no computer could ever think 25 moves ahead. Let's say there were only 3 legal moves in any given situation, or you might say there are only 3 REASONABLE moves in every situation. Than means 9 different possibilities for the first move (white has 3 moves, and for each of his moves black has 3 responses). For the first 2 moves we get a total of 9*9=81 possibilities. For the first 12 moves it is 9^12 possibilities, which happens to be a few hundred billion, just about as much as a computer can process in a reasonable amount of time, say a minute. That's why most chess programs don't think ahead more than 12 moves. But for the first 25 moves we have 9^25 possibilities, which would take any computer several billion years to go through.
I came back to edit my entry but finished too late. I admit that I'm out of my depth on this one. It's a monster topic and I just don't have the programming background to even guess at how it's done, but level 2000 hand-helds have been with us for 20 years, so it can't be all that exceptionally difficult.
Maybe it's taken this long to develop a champion-bashing program because the people doing the programming are world-class programmers - not world-class players.
I don't know. But I do know that the markets are not closed systems and are therefore not accessible by way of the same logic as are closed systems like chess games.
Quote from baggerlord:
But the computers are programmed to play book openings, so they don't have to figure any of this out. They don't start "thinking" until the game has left opening theory. And I guarantee the opening book in a word class computer is greater than any humans opening knowledge, even Kasparov.
This is an interesting thought. To defend my point, you could start at any given point in the game, I am not necessarily talking about the very first moves of the game, but you have brought up something far more intriguing: How do we know what we call opening theory is valid? We don't. In fact, the experts' views on different moves in the opening change all the time. The only reason we have opening theory (with the exception of obvious traps) is the very reason why humans are fundamentally better at chess than computers: Humans can recognize the intrinsic value of a position intuitively. Sure they are trying to express their intuition, and that's when it becomes accessible to computers: They may say the center is important or doubled pawns are bad, which are all facts that a computer could not know if they were not programmed into it just like that. If you analyze the game just 25 moves deep you will not realize how disadvantageous a doubled pawn is likely to become later on in the game. But an experienced human sees far more in a position than is even expressible in simple statements like "the center is important" or "doubled pawns are bad" etc.
My .02 is that computers are not yet capable of processing the gigantic amount of data that is going thru a trader's mind at any given time. The chess table represents a matrix with known outcomes for the current turn plus (possibly) several more. The possible outcomes of a market within a given risk tolerance is considerably higher.
Considering that it takes the average trader at least a year to become profitable, another issue is this extremely long learning curve. It's presumably difficult for a trader to tell a programmer what he wants the programmer to do and have it implemented properly. How is the programmer going to know all about market shorts, for example?
Nor can a computer easily plot (aka anticipate or plan) a move based on possible future outcomes modeled on past experiences as quickly as a human can. Por ejemplo, say Greenspan is coming on TV in 45min. How are you going to model that into a trading program? (Not that you would absolutely NEED to include something like that in a program, just making a point that certain kinds of events, which can easily be weighted and modeled in a trader's head, do not easily translate to machine code).
Plus, I have overlooked the biggest factor so far. TIME. All this data mining that AI would do in order to accurately model a live trader (getting back to the Human VS CPU chess game) doesn't occur quickly enough to compete with a fast market.
Considering that it takes the average trader at least a year to become profitable, another issue is this extremely long learning curve. It's presumably difficult for a trader to tell a programmer what he wants the programmer to do and have it implemented properly. How is the programmer going to know all about market shorts, for example?
Nor can a computer easily plot (aka anticipate or plan) a move based on possible future outcomes modeled on past experiences as quickly as a human can. Por ejemplo, say Greenspan is coming on TV in 45min. How are you going to model that into a trading program? (Not that you would absolutely NEED to include something like that in a program, just making a point that certain kinds of events, which can easily be weighted and modeled in a trader's head, do not easily translate to machine code).
Plus, I have overlooked the biggest factor so far. TIME. All this data mining that AI would do in order to accurately model a live trader (getting back to the Human VS CPU chess game) doesn't occur quickly enough to compete with a fast market.
Quote from bungrider:
I have overlooked the biggest factor so far. TIME. All this data mining that AI would do in order to accurately model a live trader (getting back to the Human VS CPU chess game) doesn't occur quickly enough to compete with a fast market.
However, a computer could easily be programmed to trade like a "disciplined" Technical Analyst; it's all math and pattern recognition...and the computer could effect TA almost instantaneously. Time would be its big advantage in this scenario. Scope too: a computer could track thousands of stocks simultaneously - something no human can do.