Equal-weight zero-cross from smoothed spreads

Agreement as a construction problem

Editorial reading: treat multi-indicator agreement as a construction problem. Put every series on one scale, measure it as a residual to its own exponential-smoothing curve, then match amplitudes before averaging. A zero-cross is then an explicit, equal-weight consensus test rather than a remembered checklist.

The archive workflow that follows is a historical construction of that kind. It does not establish present-day performance.

Four inputs and a shared 0-to-100 scale

The worked composite used four inputs: a major industrial price average; a breadth composite of net advancing issues, net up volume, and net new highs; a relative-strength reading of that breadth composite; and a running product of daily net volume and the daily net breadth composite.

Each input is normalized to a 0-to-100 scale by subtracting the series minimum from the observation, dividing by the series range, and multiplying by 100. Normalization means expressing each observation as a percentage of that series own minimum-to-maximum range so the inputs share a 0-to-100 scale.

Residuals to a 10 percent smooth

A 10 percent exponential-smoothing curve is computed on each normalized series and treated as a short-horizon support or resistance reference, described as comparable to a 20-day moving average. Exponential-smoothing is a recursively weighted short-horizon average of an ordered series; here it is the reference subtracted from each normalized input.

Each component-oscillator is the normalized series minus that exponential-smoothing curve, evaluated at every sampling interval.

Range-matching and the equal-weight average

Component oscillators are then multiplied or divided by a simple factor so they share a common amplitude. In the example, only the relative-strength residual needed adjustment because it nominally spanned about plus or minus 30 while the others spanned about plus or minus 10. Range-matching is that simple multiply-or-divide adjustment, applied before the series are averaged.

The composite-oscillator is the equal-weight average of those matched oscillators.

Zero-cross as momentum-strategy and trend-filter

A zero-cross is a change of sign in the composite. It is read as the typical series crossing its exponential-smoothing curve in the same direction and is the momentum-strategy entry or exit event. Momentum-strategy here means the full enter, exit, or stand-aside procedure that fires when the averaged residual crosses its signal level.

The sign of the average also serves as a trend-filter. The trend-filter is the sign of the averaged residual relative to a chosen threshold, used to stay with or leave the prevailing short-term direction.

Paper window, options, and overlays

The procedure was followed on paper across a 26-month window from June 1986 to September 1988. Grouped losing signals were attributed to short-cycle, volatile stretches, while trending stretches were described as the intended setting.

Optional construction changes include averaging more series to reduce composite volatility and replacing the zero threshold with staggered levels such as plus two to enter and minus two to exit.

Discretionary overlays were suggested: skip a one-day selloff signal when an advance off a major low is only two to three weeks old, and give more weight to an exit when a zero-cross coincides with a broken intermediate trendline after a multi-month advance. Editorial note: those overlays are not part of the mechanical equal-weight average.