Bollinger Bands Strategies

Bollinger Bands are a volatility envelope drawn around a moving average. The middle band is normally a 20-period simple moving average of closing prices, and the outer bands sit a multiple of the rolling standard deviation above and below it — most often two. Because that standard deviation is recomputed on every bar, the envelope widens after large moves and contracts after quiet ones. This is what the indicator actually measures: not direction, but where the current price sits relative to its own recent dispersion.

Two derived series matter as much as the bands themselves. **%b** normalises price position inside the channel — 0 at the lower band, 1 at the upper, outside that range when price closes beyond them — which makes the relationship comparable across instruments and periods. **Bandwidth (BBW)** measures the distance between the bands as a fraction of the middle band, turning the envelope into a plain volatility oscillator. Several of the versions catalogued on this page trade one of these derived series rather than the bands drawn on the chart.

## How the strategies use the bands

Most implementations belong to one of two logical families, and they are close to opposites.

The **mean-reversion** family treats a close beyond a band as a stretched condition and fades it, expecting a return towards the middle band or the opposite side. Exits are typically the moving average, the far band, or %b crossing back through a threshold. The whole family rests on the assumption that the instrument is range-bound over the holding period.

The **volatility-expansion** family reads the same chart the other way. Its clearest form is the squeeze: bandwidth compresses to a multi-month low, and the trade is taken when price then closes decisively outside the contracted range. Here the bands act as a regime detector — quiet periods tend to precede active ones — and the tradable event is the transition, not the touch.

A third, smaller family uses the bands as **structure**: the middle band as a trend filter or dynamic support, or Bollinger's own W-bottom and M-top patterns, where the second low or high is judged against the band rather than only against the previous price extreme.

## Main variants

Across the decoded versions on this page, the recurring axes of variation are:

- **Parameters and band construction** — lookback period, deviation multiple, and whether the centre line is a simple, exponential or weighted average. - **Confirmation layers** — momentum or flow oscillators (RSI-derived, stochastic, MFI), a trend-strength filter such as ADX, candlestick or price-action confirmation at the band. - **Alternative envelopes** — Keltner Channels (ATR-based rather than standard-deviation-based), linear regression bands, and other custom envelopes. Comparing an ATR envelope with a standard-deviation envelope is what defines the classic squeeze condition in the first place. - **Chart and data type** — time bars, Renko bricks, or aggregated bars, each of which feeds a different series into the standard deviation. - **Market and timeframe** — intraday FX, daily indices and equities, crypto — with the same rules behaving very differently depending on how much mean reversion the instrument's noise actually contains. - **Width-driven versions** — where the entry is derived from BBW or %b thresholds rather than from a price/band contact.

## What differentiates implementations

Two versions can share the same indicator and still be unrelated strategies. The decisive details are: the **regime assumption** (fade or follow); the **signal definition** (an intrabar touch, a close beyond the band, or a close back inside after an excursion — these are three different systems); the **directional filter**, which decides whether the logic runs everywhere or only where it is supposed to work; the **exit**, since a middle-band target, an opposite-band target, a fixed multiple of ATR and a time stop produce different trade distributions from identical entries; and the **execution convention** — signal bar close versus next bar open.

## Common mistakes

Treating a band touch as a signal in itself is the most frequent one: in a trend, price can walk along the band for many consecutive bars, and every fade taken against it is a loss. A related error is assuming the 2-deviation envelope contains a fixed statistical share of price — returns are not normally distributed, and the deviation is measured on the same recent window that produced the move.

Others: stacking several price-derived oscillators as "confirmation" when they carry largely the same information; tuning period and multiplier jointly until an isolated peak appears; ignoring that non-time bars (Renko in particular) can make historical band values unreproducible in live trading; and testing high-frequency mean-reversion rules without spreads, commissions and slippage, which those versions are especially sensitive to.

## How to evaluate a version

Start by naming the family, then test its premise directly on your instrument before testing the rules. Fix the exit logic before tuning entries, so improvements are attributable. Examine the parameter surface rather than a single setting — a broad plateau is more credible than a spike. Run the base rules with and without each filter to see what the filter actually contributes. Then push the surviving configuration through out-of-sample data, several instruments, and both calm and volatile periods, with realistic costs applied. Finally, compare against a plain baseline — a moving-average rule, a fixed-percentage envelope, or buy-and-hold over the same window — and check whether results come from the whole sample or from one concentrated stretch.

Strategies in this concept (62)

Frequently asked questions

What are the standard Bollinger Bands settings?

The conventional configuration is a 20-period simple moving average with bands at two standard deviations. It is a starting point, not a rule. The two parameters are coupled: shortening the lookback makes the deviation estimate noisier, so a shorter period usually needs a different multiple to produce comparable band excursions. If you change one, re-check the other, and prefer a setting that sits inside a broad region of similar behaviour rather than an isolated optimum.

Is a price touch of the upper or lower band a sell or buy signal?

No. A band touch only says price has moved far relative to its own recent dispersion, which happens both at exhaustion points and at the start of strong trends. During a sustained move price can close outside the band repeatedly — the 'walking the bands' condition. That is why most usable implementations add either a directional filter or a re-entry condition, such as requiring a close back inside the band, instead of trading the contact itself.

What is a Bollinger Bands squeeze?

A squeeze is a period in which bandwidth — the distance between the bands relative to the middle band — falls to a low relative to its own history, indicating unusually compressed volatility. Squeeze-based strategies do not trade the compression; they trade the expansion that follows it, entering when price breaks out of the contracted range. The classic formulation identifies the squeeze by the Bollinger Bands contracting inside the Keltner Channels.

What is the difference between Bollinger Bands and Keltner Channels?

Both are envelopes around a moving average, but they measure width differently. Bollinger Bands use the standard deviation of closing prices, which reacts sharply to sudden dispersion. Keltner Channels use Average True Range, which incorporates gaps and intrabar range and responds more smoothly. Because the two react at different speeds, their relative position is informative in itself — which is exactly what squeeze detection exploits.

Do Bollinger Bands strategies work the same on all markets and timeframes?

No, because the underlying assumption changes. Mean-reversion versions depend on the instrument oscillating around a level over the holding period, which is more typical of some FX pairs and index futures intraday than of strongly trending assets. Breakout versions depend on volatility clustering being persistent enough for the expansion to continue after entry. The same rules should be re-validated per instrument, per timeframe, and with that market's real transaction costs.

Why do decoded versions of the same indicator differ so much?

Because the indicator is only one component. Two versions can both be 'Bollinger Bands strategies' while disagreeing about direction (fade versus follow), about what constitutes a signal (touch, close beyond, close back inside), about filters, exits, position sizing and the chart type feeding the calculation. When comparing versions on this page, read those elements first — they explain the differences far better than the band settings do.

All strategy concepts · Explore all strategies