Bollinger Bands, developed by John Bollinger in the 1980s, are one of the most drawn overlays in retail charting. The construction is simple: a middle line (usually a 20-period moving average), and two outer lines placed two standard deviations above and below the middle. The result is a volatility envelope that widens when the market becomes more volatile and narrows when it becomes calmer. The information the bands provide is real but different from what most retail chartists take from them.
The mechanic
Every trading day, the standard deviation of the last 20 closes is recalculated, and the upper and lower bands are redrawn at ±2 standard deviations from the moving average. In a stable market, the bands stay close together; in a volatile market, they widen. The bands themselves are a moving picture of realized volatility.
Statistical convention says that in a normal distribution, roughly 95% of observations fall within two standard deviations of the mean. If short-term price movements were normally distributed, about 95% of prices would sit inside the bands. In practice, financial price series have fatter tails than normal distributions, so the actual coverage is somewhat less — but the "95%" heuristic is close enough to be useful as a mental frame.
What the bands describe, correctly
Range under current volatility. When price touches or breaches the upper band, it has moved unusually far above its recent average by the standards of its own recent volatility. When it touches the lower band, unusually far below. Both are descriptions of the current move's magnitude, expressed in units of the market's own recent behaviour.
Volatility regime. The width of the bands (called the "bandwidth" in some technical literature) is a visual reading of realised volatility. Long stretches of narrow bands indicate compressed volatility; sudden widening indicates a volatility expansion. These regimes have some tendency to persist, though the persistence is loose.
The squeeze. Prolonged periods of very narrow bands are often followed by directional moves once volatility expands again. The move can go either direction — the bands describe magnitude, not direction — but the observation that a volatility expansion is likely after a long compression has empirical support.
What the bands do not describe
Directional forecasts. A price touching the upper band is not a reversal signal. In strong uptrends, price can "walk" the upper band for weeks, printing new highs and re-touching the band repeatedly. Interpreting each touch as an overbought signal produces a stream of premature exits from what the tape is describing as the strongest possible move.
The reverse is equally true for the lower band in strong downtrends. A price at the lower band is not a bounce candidate; in the middle of a serious decline, it is a description of an ongoing move that may extend meaningfully further.
Whether "abnormal" means "will revert." Bollinger himself has been consistent about this in decades of writing: touching a band is a description of a statistically notable price level relative to recent history, not a signal that price will revert. Whether it does revert depends on context that the bands alone cannot describe.
Two ways the bands add information
As a filter for other analysis. Bollinger Bands work best when combined with other indicators or price-action reads that provide the directional context they lack. A price touching the lower band while breadth is deteriorating and momentum is negative is a different situation from a price touching the lower band while breadth is strong and momentum is positive. The bands describe the magnitude; the other analysis describes the direction.
As a volatility regime frame. Reading the bandwidth over time — is the market currently in a compressed-volatility regime or an expanded one — is a valuable analytical frame in its own right. It helps set expectations for position sizing (larger in low-vol regimes, smaller in high-vol regimes for the same nominal risk), and it helps interpret the character of individual moves within the regime.
The 20-period, 2-standard-deviation convention
The default settings (20 periods, 2 standard deviations) are conventions, not natural constants. Different market types and time-frames may benefit from adjustments — some traders use 20-period 2.5-standard-deviation for volatile markets, or 10-period 2-standard-deviation for shorter time frames.
The risk of tuning the parameters is over-fitting: parameters chosen to work well in past data may not generalise. The convention exists partly because it works reasonably well across most conditions, and departures from convention should be justified by specific analytical needs rather than by "I tested various settings and this one had the best backtest."
What to stop doing
Do not treat a band touch as a decision point. It is not one. Prices touch bands frequently, and the majority of touches do not precede meaningful reversals.
Do not backtest "sell at upper band, buy at lower band" as a strategy. It is one of the most tested rules in retail literature; the results, before transaction costs, are marginal at best.
Do not read the bandwidth as a predictor of the direction of the next volatility expansion. Compressed bands often precede directional moves, but the direction of the move is not predictable from the compression pattern alone.
The rule to internalise
Bollinger Bands are a picture of the current volatility envelope. They describe magnitude of moves relative to recent history and provide a visual read of the market's current volatility regime. They do not predict direction. Reading them as a description rather than as a decision rule is the difference between using them well and adding another source of premature exits to your process.
Educational content only. Not investment advice.