That in the "black box"?

First of all, expensive. Second, we increase the dimension of the problem and affects its properties, as these indices are strongly correlated. It is therefore likely that the solution will be too inaccurate and / or unstable to errors that are inevitable. And the quality of information for today is poor. To verify <representativeness, homogeneity and stability of the> information you need some understanding of the processes taking place in the economy.

The lack of quality information can be replaced by hypotheses about the contents of <black box>. Although this is where the supporters of this approach show integrity: "We do not abstract science. We are dealing with money and do not want to risk it. Therefore, to afford the hypotheses we can not>. It seems that everything solidly. But look at what lies behind the words.

It is almost obvious: not taking any hypotheses at all, you can not get any meaningful response. And in fact, supporters of the method <Black Box> implicit use of some of his hypothesis. In each case, they are easy to select. But the main postulate is formulated as follows: <There are laws governing the functioning of the market. We do not know, but we assume that they remain unchanged. Our program is just reveals these laws and use them to make decisions>.

Algorithmic Approach

At first glance, everything is great. On the hypothesis of the existence of immutable laws based all natural science. For several thousand years, scientists have only to deal with and that reveal such immutable laws, and no harm was not. This is just something and offered - replaced Scholars Program, which will do the same. Age of Cybernetics, after all. The whole problem is that such a program can not be. A person can seek unknown laws, and the machine - no. And understand this is not difficult.

Suppose there is a program, a kind of "universal Forecaster>, which works as follows. The input to the program served some data sets, and the program gives them a sequel. And if there are patterns in changing the original series, and are input rather long intervals, the program correctly predicts continued.

For simplicity, we can assume that are input data for one indicator, receiving only two values, 0 and 11. We can not exclude the option when the series of data generated by some algorithm, that is, the data are the result of some other program. Here the pattern is clearly present, and our "universal Forecaster> must operate in this case.

But this program can be written, for example, as follows. Take text <universal programmer> and touch up at the end so that the new print out the 0 whenever the "universal Forecaster> gives one, and 1 - otherwise. Working with such a series of data, our <oracle> will be wrong at each step. That is, assuming the existence of "universal forecaster>, we inevitably come to a contradiction.

From the above discussion it is clear that the program can find only those laws which <easier> than herself. But the main supporters of the pathos of the method <Black Box> is precisely the fact that the laws in the financial markets are very complicated. What then should be the complexity of the program?