Trend Calculations

The purpose of all trend identification methods is to see past the underlying noise in the market, those erratic moves that seem to be meaningless, and find the current direction of prices. Because there may be more than one trend at any one time, caused by short-term events and long-term policy, it is possible to search for th e strongest, or most dominant trend, or a minor trend that corresponds to your expected time frame. The technique that is used to uncover the right trend depends upon whether any of the trend characteristics are known. Does it have a seasonal or cyclic component. is it based on long-term monetary policy, or is it an overnight effect? The more you know about the reasons why prices trend, the better you will be able to find the most reliable calculation for separating the price direction from market noise.
Once you know that there is a fundamental relationship between data, based on measuring the properties of dependence and correlation. a formula can be found that expresses one price movement in terms of the other prices and data. The predictive qualities of these methods are best when applied to data that has been seen before. that is, prices that are within the range of historic data. Forecasting reliability decreases sharply when values are based on extrapolation outside the previous occurrences. This phenomenon will also be true of other trending methods. Because of the way we test and define the final trend calculation, it is based on the movement of historical data: when prices move to new levels, the results of the model will often deteriorate.

LINEAR REGRESSION MODEL

A linear regression, or straight-line fit, could be the basis for a simple trading strategy similar to a moving average. For example, an n-day linear regression, applied to the closing prices, could be used with the following rules:

1. Buy when the closing price moves above the forecasted value of today’s close.

2. Sell when the closing price moves below the forecasted value of today’s close.`

There is an important difference between a model based on linear regression and one founded on a moving average. There is no lag in a regression strategy. If prices continue higher at the same rate, a moving average system will initially lag behind, then increase at the same rate. The lag creates a safety zone to absorb some changes in the direction of prices, without getting stopped out. A regression model, on the other hand, identifies a change of direction sooner by measuring future movement against a straight-line projection in which the current price value has little influence. A steady price move, however, will place the fitted line right in the center of market movement, subject to frequent whipsaws. The area at which a uniform trend changes from one direction to another is a difficult case for a linear regression system and points out the need for using bands. Even with bands, the turning point of an orderly trend will appear to have much greater variance than during the direction period over the same calculation interval.