Trading is about predicting the future and your success depends on your ability to do it. For forex and cmmodities trading one of the most interesting approaches used to predict the price movement is intermerket analysis using neural networks. Its not an easy task to explain how neural networks work in simple terms but I will try my best.
Lets look first at what intermarket analysis is. Lets say we would like to predict the price of ethanol few days in the future. What does the price of ethanol depend on ? The right anwer is on many things. Since ethanol is made from corn its obvious that direction in price of corn is one factor. One of the dirty secrets of ethanol production is that the fertilizer used for corn is produced using natural gas. So there is dependency on the direction in the price of natural gas. Same logic applies to many commodities and currency pairs.
Suppose all these relationships are known, now the question is how to analyze daily data to make correct prediction. Lets say that by analyzing historical data it has been established that price of ethanol influenced 30% by the price of corn and 20% by the price of natural gas on top of other factors. Neural network consists of multiple layers of neurons lined up in rows. Every neuron is connected to any other neuron on the adjecent layer. The first row of neurons is used as input, so every neuron represents a particular data input. The outer row consists of outputs. For the sake of example suppose there are just two outputs, the positive value on one would indicate that the price of ethanol is going up and the negative value on other that the price of ethanol is going down. In order to make prediction the inputs in our case lets say that the percantage change in the price of corn and natural gas would be multiplied by the weight(lets say 0.3 and 0.2) respectively and based on the result the value would be passed to a particular other neuron on the next layer for further analysis along with other factors combined. Since our inputs are the change in price of natural gas and corn each of these also depend in turn on other input variables. Therefore in our neural network there will be neurons analyzing the trend in the price of natural gas and corn and so on. The weights are determined through so called back propagation of error. The idea is basically train the network by changing the weights until the outputs produce correct information based on the historical data. The network of course is constantly trained and retrained using fresh market data.
The approach seems to work very well for forex and commodities trading because these have clear intermarket dependencies and seem to trend much better than stocks. The accuracy seems to be very high, as much as 80%. One of the products on the market the does this is called Vantage Point. You can check their website here http://www.tradertech.com/. I never used the product, so I can't provide personal testimony. But on my opinion its worth the try. The person behind the product is Louis Mendelsohn. I've read his books, he seems to be very credible. The company has been on the market since 1979 and they are still in business.
