Constructing a second-hour session residual exponential index

Identify the second-hour window

An hourly price archive beginning in 1971 was programmed to retest a last-hour change series as a forecast input. A scan of the other clock hours identified the 11 a.m. to noon interval, then the second trading hour on that calendar, as the additional session window to construct.

Session analysis, in this construction, isolates a defined clock-hour window and compares its point change with the week's average hourly move before any smoothing. The second-hour window is the 11 a.m. to noon interval, treated as the second trading hour of that session calendar.

One stated rationale is that public orders based on the prior close tend to arrive near the opening, after which later professional trading can offset those early judgments.

Form the weekly residual

The index input is the hourly residual: the second hour's point change minus the average point change computed across all hours.

Because the series is weekly, weekly aggregation takes that week's total second-hour point change and divides it by the number of trading days in the week. The all-hours baseline is the close-to-close change from the prior week to the current week, divided by the number of trading hours in the week.

Scale, then apply the exponential update

Display scaling multiplies average points per hour by 1,000 and rounds the result to the nearest integer before the final smooth. Exponential smoothing here is a 7.4 percent exponential update applied to that scaled weekly residual because that constant approximates a 26-week average.

Forecast horizons and threshold bands

The series was evaluated at forecast horizons of one week, five weeks, 13 weeks, 26 weeks, and one year. Threshold bands placed bullish and bearish labels two-thirds of a standard deviation from the mean so that about one quarter of weeks fell in each tail.