- 1997: Volume 6, No. 2 The Fundamentals of Data for Technical Analysis

The key to profitable investing is hidden within the history of each market. It is up to the analyst to find the key from among the many false impressions and near matches. The tools at hand-your personal computer and vast market data-offer unlimited potential for formulating, testing and proving trading methods.

Many traders, not content to follow the pack into pre-set trading schemes, are searching for their own, unique paths into trading and the markets. Every market technician wants to be ahead of the pack in calling the next market turn. Here are a few tips about using data wisely to help you on your journey.

Always Use Accurate Data

It almost goes without saying that accuracy is paramount in a financial database. Of what use are the results of a study if they are based on erroneous information? Why spend thousands of dollars on software to formulate a profitable trading system if you can't trust the results? At CSI, most corrections are made within hours of the original data posting, in lightly followed markets, an error may be corrected years later. To assure that your data is as flawless as possible, we recommend you use an error detection utility. (CSI's version is QuickTrieve) at least once a week.

Use Current Data

Basing decisions on last month's charts is like picking a play from last month's theater listings. The address is right, but you have no idea what you're getting into. Every day the markets reveal more about themselves and current data is the most revealing. The basic transaction needs of setting market orders and stops require current prices, but that is not all current data can do for you. The most recent days show you current trends, tops, reversals, etc. Current data is a compass. It shows you what direction you are going in now.

Study Extensive History

Although recent data is indeed the most revealing for the current situation, this is not to say that the study of long histories is unimportant. Nothing could be farther from the truth. Wasn't it Patrick Henry who said, "I know of no way to judge the future than by the past?" Historical data shows you how a market has performed in all kinds of circumstances, one or more of which may be repeating right now. It can show you seasonal or longer-term cycles. It can be used for intermarket studies to compare how the market reacts to or how it affects other markets. Remember that no market exists in a vacuum and intermarket relationships are often critical to market movement.

There is no substitute for a long-term historical database for back-testing a trading system. In any scientific study, the sample size affects the validity (margin of error) of the result. A medical trial using just a few test subjects is not considered as reliable as a study involving a broad segment of the population. Similarly, a trading system that seems profitable over a short period (with a small sample size of trades) is not considered as reliable as a trading system that demonstrates proficiency over the long term (with a greater sample size of trades). The statistical margin of error is reduced with every new trading situation.

Historical data is particularly important for users of neural networks, which are programmed to learn as they go along. Once a neural net theory is formulated, it must "learn" about a market from its history. The results must then be applied to a "test period" that is different from the "learning period." Any overlap between learning and testing invalidates the results, so a great deal of history is required.

Use All Your Resources

Trading systems should take advantage of all the valuable data at hand, and not overlook that which is less easily processed. For example, most trading systems for futures markets ignore the impact of volume and open interest on future price. This is largely because their impact is not easily established. To correctly process the effects of volume and open interest, the developer must first recognize that they are not independent of each other.

Volume tends to be heavily related, if not correlated, to open interest. Both generally rise as the harvest (or equivalent) approaches and fall before and during the planting seasons. This relationship makes it difficult to separate the effects of each on market action. It is futile to treat largely redundant inputs in the same manner as independent variables. Your software must avoid the pitfall of basing decisions on a majority vote of several technical indicators, all of which measure the same market characteristic. Proponents of such flawed logic tend to accept large losses without a clear picture of why the odds seem to be stacked against them.

In some of our work with volume and open interest, we have combined the pair into one parameter. This transformation simplifies the analysis process and delivers a possible reduction in parameter count. The result is a very distinct advantage. When combining volume and open interest into a single parameter, it pays to analyze what volume measures. If open interest increases (or decreases) on a given day, volume, by definition, must rise a minimum of the same quantity to support the open interest growth (or drop). Any additional volume above that prompted by a change in open interest can be attributed to influences that are above and beyond the effects of open interest movement. Volume and open interest can supply useful information, but only if your chosen trading system meets the challenge of finding it.

Bob Pelletier is president of Commodity Systems Inc., a data retrieval service offering daily updates and end-of-day historical data on commodities, stocks, options, indexes and mutual funds. In addition to their data feed, they provide software to assist traders with data retrieval and maintenance, graphic analysis, accounting and trading system evaluation. For more information, contact CSI at (561) 392-8663.