Who this entry is for — The moving average is the world's most used indicator — and almost always drawn in the wrong place. Hurst explains what it really is (a cycle filter), why 10 and 30 weeks work, and where it belongs on the chart.
Source: J. M. Hurst, The Profit Magic of Stock Transaction Timing, Prentice-Hall, 1970 — Chapter 3, The Significance of Moving Averages → How a Moving Average Can Aid Cyclic Analysis (pp. 62–66, Figs. III-10/III-11). Mathematical derivation: Appendix IV.
Prerequisites
Hurst nominal cycles — you need the duration table to see where the "cutoffs" fall.
What a moving average really does
In plain words — It is a smoother: it kills the short waves and lets the long ones through. And the span — how many bars you average — is the knob that decides which waves to kill.
A moving average is a numerical process applied to a sequence of prices that reduces short fluctuations while leaving the long ones nearly untouched. The key results (derived in Appendix IV, summarized here):
- an average of span N exactly zeroes fluctuations of duration equal to N;
- it greatly reduces (without necessarily eliminating) all those of duration less than N;
- it lets through the longer ones: heavily attenuated just above the cutoff, the very long nearly unscathed.
The cutoff is therefore a design variable: choosing the span decides which periodicities to suppress and which to observe.
Why 10 and 30 weeks, of all numbers
In plain words — The era's two most popular averages fall, of all places, exactly between the model's nominal cycles: the 10 between 6.5 and 13 weeks, the 30 between 26 and 39. Charting found them by trial and error; the model explains why they work.
| Average | The cutoff falls… | Effect |
|---|---|---|
| 10 weeks | between the 6.5 and 13 nominals | shows the 13-week-and-longer cycles, kills the shorter |
| 30 weeks | between the 26 and 39 (9-month) nominals | shows the 9-month-and-longer, kills 26 and below |
"If we had set up these smoothing objectives in advance with full knowledge of the model," Hurst writes, "we could not have chosen the cutoff points much better." The chartists, empirically, had found the right answers.
The proper placement: half a span back
In plain words — The average of the last 10 weeks belongs to the middle of those 10 weeks, not to the last one. Plotted on the last bar, the chart lies; shifted back 5½ weeks, the average runs down the middle of the channel.
It is the "small but important" characteristic charting had overlooked: the average value computed on the last N prices should be associated with the price half a span back — for the 10-week, halfway between the fifth and sixth previous data points. (Hence the practical advice: use an odd span, so every average value has a bar to belong to.)
On Alloys Unlimited the effect is stark: shifted 5½ weeks, the 10-week "represents a smoothing of actual price fluctuations" and runs "almost precisely down the center" of one of the constant-width channels. The 13-week cycle just barely comes through (the little dip at the top of the May–June peak); the ~24-week component and everything longer are "tracked to perfection".
The two cyclic uses
- Isolating a component — by picking the span, price oscillates about the average in sympathy with the shortest component not suppressed: placing the envelope correctly, with the right highs and lows, becomes unambiguous.
- Estimating the present — the centred average stops half a span before the last data point, but that missing half-cycle can be estimated "quite accurately" from the remaining price motion — and the envelope with it.
Warning — "The average lags" is the wrong complaint: plotted where it belongs, the average chases nothing — it describes, with half a span of modesty. The real cost is that one: the last N/2 points do not exist yet. How to live with that operationally is Chapters 4–6 material (half-span and projections).
Summary card
| Fact | Value |
|---|---|
| Average of span N | Zeroes cycles of duration N, attenuates shorter ones |
| 10-week cutoff | Between the 6.5 and 13 nominals |
| 30-week cutoff | Between the 26 and 39 (9-month) nominals |
| Correct placement | Half a span back (10 wk → −5½) |
| Recommended span | Odd, so each value has its bar |
| Cyclic uses | Isolate components · place envelopes · estimate the present |
Links
- Hurst nominal cycles — where the cutoffs fall
- Half-span and full-span — the operational development (Ch. 6)
- Chart pattern verification — Chapter 3's framework
- Hurst tradition — chapter index