Trading Commodities Market

Raymond Raud

July, 1997

Forecasting Topics Knowledge in Price Data Modeling with Neural Networks Timing Tools and Methods

Trading futures market is simple." Buy low and Sell high" is the only rule to follow. The winning trick is to know the market turning points before other traders. The commodity market is a "zero sum game", there are precisely as many buyers as sellers, same amount of money lost as won.

The market moves (price changes) only due to the actions of traders. Traders are influenced by their own past experience, by the past experience of the people they believe (analysts, brokers, friends), new events, assumptions of other traders behavior. Traders make the market turn. To win you always have to know what the majority of traders are up before they know it themselves. That sounds like an impossible proposition, but it is not. The key here is "the majority". One does not need to know how each trader will react, but only how most of them will. That is possible with reasonably high probability, knowing how most traders reacted to the similar situation in the past.

Clearly, no matter how good you or your forecasting system are, it does not work always. Hence, besides a winning forecasting method (or system) one also needs a matching trading strategy that uses strong aspects of the forecasting method and compensates for the weaknesses. In addition, a prudent money management strategy must be in place that complements both of the above.

The market reaction to the situation, its behavior, is captured in the price movement data. That is why most of the market timing tools and methods use some form of price data analysis. Besides previous period's price movement traders are influenced by the current events, relevant to their market. That is the other input for the market. We'll see the reaction to the events in the price data, but that will be already the next cycle. It is lost for forecasting. Following illustrates the rough information model of the market.

External Events for each market are different, their influence to the market varies in time, and it is difficult to find reliable historical data on the events. Considering the practical goal of forecasting the critical questions are:

Is forecasting from price data accurate enough for profitable trading?
How to build a forecasting system with highest accuracy using the readily avaialble price data?

These are the questions I addressed in my research. Your questions and discussion on any of my conclusions and assumptions are welcomed.
Please drop me an e-mail, thank you.
Raymond Raud