1998issue C111-6
Regression channels anchored to Williams %R turning windows
A regression channel sets its slope and width on a hand-drawn range, then is judged by whether later price still respects the projected rails. This note uses Williams %R turning dates as those start and end anchors instead of picking them by eye.
- A regression channel fits a linear-regression midline to a chosen window and places parallel price-channel rails at the farthest price from that line.
- A 14-period Williams %R window can replace eye-picked dates so the start and end anchors of the hand-drawn range are repeatable.
- The channel is kept only while later price in the predictive range still respects the same projected rails.
- After a channel bottom breaks, that same rail can later act as resistance at an impact zone.
How the channel is built
A regression channel is built by fitting a linear-regression midline over a user-chosen lookback and placing parallel rails at the farthest price from that line. Those outer rails are a price channel: they bound a trend's swing extremes and can be projected forward as a testable path.
The hand-drawn range is the inclusive start-to-end window that sets slope and width. By construction, no price inside that window sits outside the channel.
Choosing start and end dates by eye is subjective. Two analysts can place different channels on the same chart.
Oscillator dates as start and end anchors
Williams %R is used here only to mark overbought and oversold dates that can serve as channel start and end anchors. In this method an uptrend window is anchored at a 14-period Williams %R reading of -80 or lower and closed at a reading of -20 or higher. A downtrend window reverses that pair.
Sideways price can use either oscillator pairing, with only a slight preference for the oversold-start and overbought-end pairing.
A 14-period Williams %R setting was kept after informal trials, including Fibonacci-length averages, because it matched or beat the alternatives and was already the software default.
When existing channels were re-anchored to those oscillator dates, the author reports improved later containment in nearly every adjusted case and less trial-and-error placement.
Testing the projected rails
A channel is treated as valid when already-observed turns sit on the rails and later price continues to respect those same projected rails until a close beyond the range ends the trend reading.
The predictive range is the later segment under the extended rails. It is the window used to check whether the channel still contains later price. A historical channel is called moot when its slope keeps those rails away from any reachable later price.
Once a channel bottom is broken, the same rail can later act as resistance at subsequent impact zones. An impact zone is a later date cluster where price meets an earlier channel rail and the support or resistance reading can be confirmed or rejected.
All readings on this track · 13 readings
- 1987Constructing a volume-confirmed Williams %R
- 1991Audit inverse-range oscillators before stacking stochastic %K and Williams %R
- 1991Signed midpoint range oscillator from stochastic and Williams
- 1993Confirm an intradate candlestick only after a longer cycle reprints it
- 1994Building average directional index, the stochastic pair, and Williams percent R from highs, lows, and closes
- 1994Label the tape before you read stochastic or Williams %R
- 1996Calibrating Williams %R entries in rising channels
- 1997Dynamic zones for oscillator buy and sell levels
- 1998Regression channels anchored to Williams %R turning windows
- 1999Constructing isolated synthetic waveforms to watch indicator settling
- 2000Choosing a scale for moving-average oscillators
- 2004Splitting entry and exit speed by regime
- 2008Count the run, then confirm the pivot at a channel edge