1999issue C031-8
Combination filters with Bollinger Bands and the relative strength index
Bollinger Bands mark volatility-adjusted extremes around a moving average. A combination-filter uses those extremes only when the relative strength index confirms them.
- Fixed-percentage envelopes keep a constant width, so the channel does not change when volatility changes.
- Bollinger Bands set standard-deviation-bandwidth from the observed price series, which produces squeeze-and-expansion as volatility falls and rises.
- A combination-filter records an entry or exit only when a band extreme and a relative-strength-index confirmation both appear.
- A two-sided stack, a long-only-variant, and a buy-and-hold-baseline belong in the same comparison. A poor result on one sample does not dismiss the design.
Two stages, not one line
An envelope is a three-line price channel built from a central moving average plus upper and lower boundaries that mark temporary extremes around the trend. A combination-filter does not treat those extremes as complete signals. It waits for a second, independent check from the relative strength index, an oscillator that measures the speed and magnitude of recent price changes.
Fixed width versus adaptive width
Fixed-percentage envelopes place the upper and lower lines a constant fraction of a moving average away from that average. Channel width then stays the same when volatility changes.
Bollinger Bands replace that fixed fraction with a multiple of the standard deviation of price. Bandwidth is then tied to the observed price series rather than to an arbitrary percentage. That construction is standard-deviation-bandwidth. A common setting uses two standard deviations because, in a normal distribution, most values fall within two standard deviations of the mean. From that fact the archive infers that most of a stock's price action should stay between the bands.
On one historical sample, the two-standard-deviation bands contained more of the price bars than a fixed-percentage envelope around a simple moving average, including during sharp moves.
Squeeze-and-expansion and input choice
The bands widen when volatility is high and contract when volatility is low. That squeeze-and-expansion is the visible difference from a fixed envelope. On one sample the bands were much wider in a volatile week than during a quieter stretch.
The moving-average length is chosen to match the trend horizon being followed. The bands may be computed from closes or from alternative inputs such as typical-price or weighted-close. Typical-price is formed from the high, low, and close of each bar. Weighted-close weights the close twice as heavily as the high and the low.
Confirming the extreme
The archive combines Bollinger Bands with the relative strength index so that band extremes are used only when the oscillator confirms. That rule set is the combination-filter. It was written in explicit long-and-short form and in a long-only-variant that keeps the same logic but restricts entries and exits to the long side.
Those variants were placed next to a buy-and-hold-baseline, an explicit comparison path that stays fully invested over the same sample window. The tested rules included no stop-loss provisions. The archive treats that omission as a reason a practical system would have exited a large losing short well before the drawdown ran its course.
How to use a comparison window
A combination that fails on one stock and window is not thereby dismissed. The archive's conclusion is that usefulness depends on market selection, historical testing, money management, and the trader's style, because no method works for all stocks in all conditions.
Editorial reading: keep the two-sided rules, the long-only-variant, and the buy-and-hold-baseline in one comparison so the stack can be inspected. The comparison is a design check, not a contest to name a winner.
All readings on this track · 45 readings
- 1992Constructing volatility-scaled bands with relative strength index confirmation
- 1994Implied volatility as a band-defined regime filter for index options
- 1995Constructing projection bands from least-squares slopes
- 1995Constructing regression projection bands and range oscillators
- 1996Constructing Bollinger bands, percent-b, and stochastics
- 1996Constructing mechanical rules from Bollinger Bands and stochastics
- 1996Constructing a standard-error envelope around a linear regression
- 1996Dual-horizon ratio envelopes and regression error channels
- 1997Rational group structure with a trend screen, RSI, and bands
- 1997Asymmetric volatility band construction
- 1998Constructing three-state filters from Bollinger band envelopes
- 1999Combination filters with Bollinger Bands and the relative strength index
- 1999Constructing stochastic timed exits and band-RSI reversals
- 1999Evaluating Bollinger Bands against fixed-width and range-based envelopes
- 2000Constructing a Bollinger Band target as a forward price
- 2001Numeric candlestick encoding with local size bands
- 2001Ranked candlestick sentiment to band-cross entries
- 2002Combining Bollinger Bands, RSI, and a stop-loss
- 2002Bollinger Bands remain filters, not forecasts
- 2002Constructing a stochastic RSI with Bollinger bands
- 2002Constructing a StochRSI and Bollinger mechanical system
- 2003Constructing volatility-scaled Bollinger envelopes
- 2003Why tick breadth fails as a market personality
- 2005Constructing Bollinger bands versus fixed trading bands
- 2006Squared versus absolute deviation in envelope construction
- 2006Confirming yen crossovers with implied volatility and bands
- 2006A daily candle reversal is a hypothesis until shorter sessions fail at the same zone
- 2008Rebuild the Relative Strength Index as price-scale bands
- 2008Reading Relative Strength Index extremes on one price axis with Bollinger Bands and moving averages
- 2011Three-filter confirmation for short-swing futures
- 2011Constructing an inverse Fisher stochastic with bands and averages
- 2012Constructing a Bollinger Band indicator suite
- 2012Stacking price extremes, crossovers, bands, and MACD
- 2012Adaptive Bollinger band impulse, trend, and momentum filters
- 2013Rescaling stochastic, percent-B, and wave-count parameters
- 2014Industry-group quartile pivots as a Bollinger Bands case study
- 2014Bollinger Bands as adaptive price envelopes: a 2014 classroom case
- 2016Trend-channel entry rules from stacked moving averages
- 2016A permission stack for Bollinger, RSI, and the 50-period average
- 2017Constructing weighted Bollinger bands and volume averages
- 2017Four swing-entry rules that share a timed exit
- 2017Two-wave monthly cycles as a regime filter
- 2019Constructing exponential-deviation-bands from a midline-average
- 2020Critiquing exponential variants of Bollinger Bands
- 2020Constructing selectable volatility and moving-average bands