2014issue C0116-25
From aliasing and dilation to a roofed stochastic swing
Daily bars remain one sample per session, so aliasing several octaves below the two-bar Nyquist period swamps four-bar and eight-bar cycle estimates, and cyclic amplitude grows about 6 dB per octave. A two-pole smoother stacked with a second-order high-pass roofs that series; a later stochastic becomes a swing rule only after confirmation delay is counted against the dominant-cycle half-swing.
- Daily bars remain one sample per session even when the high and low are averaged in, so aliasing several octaves below the two-bar Nyquist period swamps four-bar and eight-bar cycle estimates.
- Simple and exponential averages buy stopband only by adding length and lag; a two-pole smoother plus a second-order high-pass roofs aliasing and six-decibel-per-octave dilation before the stochastic is formed.
- A confirmation stochastic that recrosses 20 or 80 can spend about eight bars of delay against a ten-bar half-swing of a twenty-bar dominant cycle.
- Over about 2,500 daily samples spanning 10 years the roofed stochastic's value density was nearly that of a sine wave, which historically motivated an early 20/80 cross or a swing-setup that pairs a recent cycle trough with still-falling momentum.
The archive construction begins with the sampling interval, then a roofing preprocessor, then a stochastic oscillator computed on that roofed series and smoothed once. A confirmation swing on that stochastic is then counted against the half-swing of an assumed dominant cycle.
Aliasing below the Nyquist period
Daily bars remain one sample per session even if the high and low are averaged in. Aliasing-noise is spurious short-period energy created because those bars are discrete samples, not a continuous path, so cycles near and below the two-bar Nyquist period cannot be recovered. Aliasing several octaves below that two-bar Nyquist period swamps four-bar and eight-bar cycle estimates.
Inefficient averages
Simple and exponential averages reduce aliasing only by lengthening the window, which adds lag, so they are inefficient stopband tools for sampled price series. With both critical periods set to 10 bars, a 10-bar exponential average shows only 13 dB of rejection at the Nyquist frequency. A two-pole smoother, converted from a two-pole analog prototype, is built to nearly remove that aliasing with less lag.
Spectral dilation
Cyclic amplitude grows about 6 dB per octave of period. That spectral dilation lets longer waves dominate raw oscillator output. First-order oscillators such as the stochastic, momentum, relative-strength, moving-average-convergence, and commodity-channel constructions roll off at the same 6 dB per octave and can at best flatten that dilated spectrum.
A second-order high-pass
A first-order high-pass leaves a non-zero mean in a long uptrend because longer cycles leak. A second-order high-pass at 12 dB per octave exceeds the 6 dB dilation slope and recenters the oscillator near zero with less leak-induced lag.
Roofing then one stochastic smooth
Stacking the second-order high-pass with the two-pole smoother yields a roofing filter. The stacked preprocessor is what later oscillators see: a band-limited, near-zero-mean series. A following stochastic is the current roofing value minus its lookback low, divided by the lookback range, then smoothed once instead of with stacked conventional smoothers.
Confirmation delay versus the half-swing
Confirmation-style stochastic swing rules buy after a cross back above 20 and sell short after a cross back below 80. About two bars of smoother lag, one bar of next-bar entry, and about three bars of confirmation total about eight bars of delay. That delay is set against an assumed 20-bar dominant cycle whose half-swing is 10 bars.
Early crosses and the swing-setup
Over about 2,500 daily samples spanning 10 years, the roofing-filtered stochastic's value density is nearly that of a sine wave. That historical density motivates acting when the line crosses under 20 or over 80 before the extremum, replacing confirmation delay with about two bars of net lag versus price turning points.
A swing-setup procedure pairs a smoothed, zero-mean cycle line that stays nearly in step with price peaks and troughs with a lagged stochastic-like momentum line. A candidate long setup is a cycle trough in the last few days while momentum is still declining or at a minimum.
All readings on this track · 35 readings
- 2001Swing trading with trailing stops and fixed loss exits
- 2003Linear regression swing alerts and trailing stops
- 2005Range-width gates for swing entries at base edges
- 2006Beyond setups: a six-factor trading process
- 2007Angle of ascent, chart scale, and style-fit
- 2010Stacking a short-horizon oscillator with a moving average and swing hold rules
- 2010A weekly contest long treated as one swing procedure
- 2011Four-color volume-price states for long-only swing actions
- 2012Classify momentum, velocity, or volatility before the swing signal
- 2013Swing rules as a portable entry and exit procedure
- 2013Constructing paired percent-b and stochastic swing oscillators
- 2013Give the fast band line and the slow stochastic complementary jobs
- 2013Smoothed percent-b divergences for numbered swing-wave rules
- 2013Portable swing rules for candle turns and a 1-2-3 wave count
- 2013A three-average swing and breakout case study
- 2013Constructing swing trades with a fifty-day average and a five-bar exponential average
- 2014From aliasing and dilation to a roofed stochastic swing
- 2014Channel width and trend-filter lookback as separate swing-trading choices
- 2016Constructing volume-confirmed multiday breakout swings
- 2017Constructing swing signals from a volume-weighted average cross
- 2017Golden-cross swing entry with a trailing-stop exit
- 2017One swing procedure: group leadership first, then a volume-price leave
- 2017Four-day green-candle breakout entry as one swing procedure
- 2017Swing pivots need a 100-period crossover and a breakout candle
- 2017Supply and demand zones as bounded swing maps
- 2018Three-bar volume-confirmed swing breakout
- 2018Tight stops first on swing breakouts with a range filter
- 2018Constructing trend and swing rules from horizon to crossover
- 2019Decade-level breakouts as one closed swing procedure
- 2019Inverse-pair swing trading with a shared buy-stop
- 2019Week-range next-day breakout on bitcoin pairs
- 2019Ranking strength before confirming a multi-timeframe breakout
- 2019Forty-five-degree multi-week swing breakout as a closed procedure
- 2020A dual pending-stop pair as one bitcoin swing construction
- 2020High-low activator, directional oscillator, and momentum swing agreement