How to Compare Band Indicators: The Containment Test

Search for a Bollinger Bands® alternative and you'll find the same shortlist everywhere: Keltner Channels, Donchian Channels, STARC bands. What you'll rarely find is a way to decide between them. Most comparisons describe how each band is built and leave the judgment to you.

This article takes a different route. It explains why traders go looking for an alternative in the first place, what the usual candidates actually change, and — most importantly — the one question that lets you compare any two bands on evidence rather than aesthetics: how often does price actually stay inside the band?

Why traders look for an alternative

Bollinger Bands® are built from price levels: a simple moving average of price, with bands set a number of standard deviations of price above and below it. That construction is simple and familiar, but it has a known behavior that frustrates many users.

In a sustained trend, a moving average of price lags. The band inherits that lag: it sits off-center behind the move, price rides the upper or lower band for weeks, and the band's width reflects where price was rather than how it has recently been moving. Traders describe the symptoms in different ways — "price walks the band," "the bands balloon after the move," "everything looks overbought" — but the cause is the same: the statistics are computed on raw price levels, and price levels trend.

That's usually the moment someone starts searching for an alternative.

The usual candidates, and what they change

The common alternatives each swap out one ingredient:

Keltner Channels replace the standard-deviation width with Average True Range (ATR), typically around an exponential moving average. ATR is smoother than a price standard deviation, so the bands react less to single-bar outliers — but the channel is still centered on a moving average of price, so the lag-in-trend behavior remains.

Donchian Channels drop statistics entirely and plot the highest high and lowest low over a lookback window. Simple and robust, but a single extreme bar sets the boundary for the entire lookback window, and the band only releases it once that bar ages out.

STARC bands combine a short moving average with an ATR-based width — a hybrid of the two ideas above.

All three are reasonable tools, and each changes how the band reacts. But none of them changes what the band is measured on, and none is typically published with evidence on the question that actually distinguishes bands from one another.

The question that separates bands: containment

Every band on a chart is making an implicit statistical claim: price should usually stay inside this. That claim is testable. You can take any band, run it over historical data, and count how often the next close actually landed inside it. That number is the band's containment rate, and it turns "which band is better?" from a matter of taste into a measurement.

It also reveals something most comparisons skip: a band can only be too wide, too narrow, or calibrated. A band that contains 99% of closes tells you almost nothing when price touches it; a band that contains 50% is noise. What you want is a band whose observed containment matches what its construction implies — consistently, across markets and time, not just on one backtest window.

This is exactly the test documented in Oisigma's working paper and summarized on the Proof page.

What a return-space band does differently

The Behavioral Transform Model (BTM) — Oisigma's calibrated expected-range indicator for TradingView — starts from a different design choice than every band above: it measures behavior in return space, not price space. At each bar it looks at the last ~60 bars of percentage moves, estimates an expected price anchored to the prior close, and projects a range from the dispersion of those returns. (The full logic is on the How It Works page.)

Because returns don't trend the way price levels do, the range stays centered on where price actually is and scales to current conditions, instead of lagging behind a trend or ballooning after it.

The difference shows up directly in the containment numbers measured in the working paper:

• On daily S&P 500 data from 1928 to 2024 — about 24,000 trading days — the next close landed inside BTM's inner band 71.2% of the time, and that rate stayed within a narrow 68.7–73.7% band in every calendar decade, through the Great Depression, the 1987 crash, 2008, and COVID.
• Across 40 instruments in five asset classes — equities, FX, commodities, rates, and crypto — the cross-instrument average was ~71.6% for the inner band and ~94% for the outer band.
• Head-to-head on the same 40-instrument universe, with both bands at ±1 standard deviation, BTM contained the next close ~71.65% of the time versus ~42% for a standard 20-bar Bollinger Band® at ±1 standard deviations. At ±2 standard deviations — the canonical (20, 2) Bollinger default — the comparison was ~94% versus ~83%. BTM contained price more reliably on every one of the 40 instruments, at both bands. And with BTM's lookback forced onto Bollinger's own 20 bars, the return-space band still held ~69% at ±1σ — the working paper attributes only about 3 points of the ~29-point gap to the window; the rest is the return-space formulation.

All of these figures are historical calibration results measured in research. Past behavior is not a guarantee of future results.

The paper is equally direct about the model's limits: calibration is an average property, not a moment-by-moment one. In the first days of a fast crisis, when volatility spikes faster than a rolling window can register, containment historically fell to roughly 65%. A band that publishes that number is telling you how to interpret it honestly.

What no band can do

One thing doesn't change no matter which alternative you pick: no band predicts where price will go. A band describes where price is relative to its own recent behavior — inside its normal range or beyond it. That's context for your own analysis, not a signal, and touching a band edge is not a forecast of reversal. Any comparison that promises otherwise is comparing marketing, not methods.

BTM is deliberately positioned this way: no arrows, no entries, no buy/sell calls — an expected range, recalculated every bar, with the methodology and every figure published for you to check.

Choosing on evidence

If you're weighing a Bollinger Bands® alternative, the practical checklist is short: ask what the band is measured on (price levels or returns), ask what containment rate its construction implies, and ask whether anyone has actually measured that rate across markets and decades — and published the misses along with the hits. For the side-by-side chart comparison, see Oisigma's BTM vs Bollinger Bands® page.

The best way to evaluate a band, though, is on the markets you actually trade. BTM comes with a free 30-day trial on TradingView — watch the range recalculate on your own symbols and timeframes, and read the evidence alongside it.

Start your free 30-day trial →