Thursday, September 13, 2012

The diversification of parameters. The portfolio of strategies.

The activities on derivatives market, especially on Forex, are associated with risk. The risk that is inherent in any operation we do. Of course, we try to minimize them, at least to control, maintain an acceptable, a predetermined level. One of the tools we use for this purpose is broadly defined, diversification.



Common understanding of diversification is often based on a simple, almost naive scheme: diversity typically includes: various types of financial instruments, often a series of futures contracts with different expiration times. Sometimes they apply to different types of goods, on which the contracts are based on: metals, agricultural products, luxury goods, antiques. And finally there are also geographical differences: East-West, developing-developed economy, etc..

But there is another concept of diversification, which is based on algorithmic trading systems: such systems are based on mathematical intelligent calculations, often rely on estimation and prediction of numerical attributes. We are dealing with probability distributions for futures or currency pairs. However, the methods of estimation and forecasting require setting the parameters of their operation. These parameters ultimately determine the action of the transaction system.

How to select these parameters? Define arbitrarily? Set manually? Here comes the concept of building a complex strategy, which is a parallel implementation of many duplicated elementary strategies, which differ only in the values ​​of these parameters.

Such a complex strategy can operate on a single currency pair and still provide risk reduction. This approach, for my purpose, I refer to as the diversification of the parameters. At the same time a set of such strategies with different parameter values is named portfolio of strategies.

This concept is not my original idea, but I'm trying to develop it extensively, to invent complex transaction systems, including systems based on combinations of elementary ones. I found inspiration while reading an excellent book by Ralph Vince, "The Mathematics of Money Management: Risk Analysis Techniques for Traders".

The idea is described, just in the context of the theory of Markovitz, in the following paragraph. Let me quote it below:


If we are measuring the correlation of prices, what if we have two systems on the same market? (...) We want to measure the correlations of daily equity changes between the different market systems.

In subsequent posts I will try to provide specific elementary algorithms. On them will be built later complex systems.


4 comments:

  1. Diversification of financial instruments is not naive if you first measure the correlation between their prices.

    ReplyDelete
    Replies
    1. And the same applies to the correlation between the results obtained by the different strategies for the same instrument.

      Delete
  2. Hello Michael,

    I did a post on my blog on the exact topic, but I called it parameter insensitive models. These type of strategies offer risk reduction and are more stable over time.

    Mike

    ReplyDelete
    Replies
    1. Also, I plan to get rid of the parameters of my strategies through the use of empirical optimization. For this purpose, the method of regularization will be applied. Although it also requires explicit setting of parameters, I hope that it will be insensitive to their values.

      Delete