Private information is practically the source of every large
modern fortune.
—An Ideal Husband, Oscar Wilde

3.1 INTRODUCTION

he discussion in Chapter 2 is couched largely in terms of trading rules based on estimates of spread ranges calculated on moving windows of data history. Figure 3.1 shows the bands calculated as mean plus or minus one standard deviation using a window of 60 days for the CAL–AMR spread. (Compare this with Figure 2.4, wherein the limits are calculated using the maximum −20 percent range and minimum +20 percent range over a 60-day window, and review the discussion in Section 2.2.) Implicit in these trading rules is a forecast that the spread will in the near future return to the local mean. Figure 3.2 shows the CAL–AMR spread again, this time with the implied forecast function.

In formal terms the point forecast, or expected value, for every future time period is the currently estimated mean value. Now it is not really believed that the spread will actually equal the mean each time period, or even one time period in the near future. (Obviously the trading rules anticipate systematic variation above and below the mean.) It is simply that the best guess based on using the moving average model is that in the near future, the spread will likely exhibit values centered on the mean. How near is ‘‘near’’ remains, like so much else, unspecified. ...