Immediate definition
In plain words — Hurst adds a few calculations to the chart to address three questions: how much of a move may remain, where the channel centre lies, and which short oscillations are hidden beneath the slower trend.
Chapter 6 of The Profit Magic of Stock Transaction Timing does not present a self-contained algorithm. It connects three tools: the half-span moving average, the full-span moving average, and the inverse moving average. All depend on the cycle duration selected at the start. If that estimate changes, their spans, lines, crossing, and target zone change too.
From cycle to estimate in four steps
The process starts with an observed duration T, estimated through cycle and
envelope analysis. The half-span window is T/2; the full-span window is T.
Because both averages are centred, they stop before the latest bar and their
final sections must be extended graphically.
When the half-span reverses, Hurst looks for the crossing C of price and the
extended lines. Within his model, the distance between the previous extreme
and C represents half the leg. The same distance is projected beyond C.
A zone of ±10% of the estimated total move surrounds the target. The
theoretical time of the extreme is the crossing plus half the half-span,
approximately T/4.
What the three tools actually do
The half-span measures a leg using the symmetry
assumed by the model. The full-span attenuates the component with duration T
and is interpreted as an approximate centre line of its channel. Its slope
may suggest that the slower channel has already turned even while price highs
and lows remain ambiguous.
The inverse moving average is price minus a centred moving average. It does not invert a forecast; it removes the smoothed part and leaves a residual around zero. In the Alloys Unlimited case, Hurst uses an odd 11-week span and reads a component near 12.7 weeks in the residual, with an observed 10–16-week range and roughly seven points of peak-to-peak movement.
The right-edge problem
A centred average needs observations that occur after the time to which its value is assigned. Complete lines and the inverse therefore do not extend to today. The historical method fills some of that gap with graphical extrapolation: a plausible but unobserved segment that may change as new data arrive. Modern software should distinguish calculated values, estimates, and unavailable sections.
The ±10% zone is also Hurst's operating tolerance, not a statistical confidence interval. All four extremes in his selected example enter their zones, while the timing estimates show substantial error. The case explains the mechanics but cannot establish their general reliability.
How it should be assessed today
A contemporary test should predefine how T is estimated, how a reversal is
confirmed, and how the right edge is extended. It should include every signal,
cost, delay, and failed case, and separate model development from out-of-sample
evaluation. Research on backtest overfitting explains why selecting settings
after many trials can yield fragile results. It does not test Hurst directly;
it supplies the relevant validation standard.
The source-checked label means that this account was compared with the 1970
text. It does not certify profitable or reliable present-day forecasts.
Sources
- J. M. Hurst, The Profit Magic of Stock Transaction Timing, Prentice-Hall, 1970, ch. 6, pp. 97–113 and appendix IV — historical procedure, example, and moving-average response.
- NIST/SEMATECH, Centered Moving Average — window centring and treatment of even spans.
- David H. Bailey et al., The Probability of Backtest Overfitting — selection risk after trying many configurations; not a test of Hurst's method.
Related entries
- Half-span and full-span — formula, centring, and the edge.
- Inverse moving average — the filtered residual.
- Case Alloys Unlimited — the four estimates in context.
- Cyclic moving averages — the Chapter 3 foundation.