Bollinger Squeezes and False Breakouts: Why an Outside Close May Not Clear Resistance

A close above the upper Bollinger Band can look like a breakout while the price remains inside its old trading range. The apparent contradiction disappears once the boundaries are separated: the band measures deviation from a recent average; a previous high records a price the market actually reached.

A new close also changes the average and standard deviation used to draw the bands. Breaking a moving statistical boundary is therefore not equivalent to clearing a previously identified resistance zone. That distinction matters most when narrow bands start to expand.

Bandwidth and %b answer different questions

A common configuration uses a 20-period simple moving average, with bands two standard deviations above and below it. These are defaults, not universally optimal settings. For the examples here, the input is closing price and standard deviation is calculated as a population measure: average the squared deviations of all 20 observations, then take the square root.

Percentage bandwidth = (upper band − lower band) ÷ middle band × 100. A middle band of 100 and standard deviation of 1 produce bands at 102 and 98, hence bandwidth of 4%. With the same mean and deviation of 0.5, the bands become 101 and 99 and bandwidth falls to 2%.

The position measure %b = (latest price − lower band) ÷ (upper band − lower band). With bands at 101 and 99, a price of 100.5 produces 0.75. A value of 1 is at the upper band; above 1 is outside it; below 0 is below the lower band. Position describes relative price location, while bandwidth describes relative dispersion. Neither by itself fixes the next direction.

Boundary cases cannot be ignored. Identical input prices collapse the bands, making the %b denominator zero. A middle band close to zero or negative can make percentage bandwidth misleading. Sample rather than population standard deviation, or another price input, also changes the values. Align those conventions before comparing charts.

How narrow is narrow?

A 2% bandwidth might be ordinary for a quiet stock and unusually low for a more volatile instrument. A universal 2% or 5% threshold can therefore miss one kind of compression while overstating another. The relevant comparison uses the same instrument, timeframe and data treatment.

One reproducible definition compares today's reading with the preceding 120 completed readings. If the current bandwidth is 2% and six earlier values are strictly smaller, the fraction below it is 6 ÷ 120 = 5%. This is a ranking rule, not a proven optimum. Whether to include the current reading, how to treat ties and how long a history to use must be fixed in advance.

A low rank supplies neither a countdown nor a direction. Compression can persist or return after a brief expansion. Intraday seasonality matters too: comparing lunchtime with the busiest minutes may identify a routine activity pattern rather than a fresh trading opportunity.

A complete example: outside the band, below resistance

Suppose the oldest ten of the latest 20 closes are all 99.5 and the newest ten are all 100.5. The mean is 100. Every squared deviation is 0.25, so population standard deviation is 0.5. Upper and lower bands are 101 and 99, bandwidth is 2%, and the latest 100.5 gives %b of 0.75.

Separately assume that an intrabar wick during this interval reached 103.5. That previous high is above the upper band because the bands in this example use closes, not the intrabar high. Now a new bar closes at 103, removing the oldest 99.5 from the rolling window.

The new window has nine observations at 99.5, ten at 100.5 and one at 103. Their sum is 2,003.5, giving a mean of 100.175. The sum of squared prices is 200,713.75; divide by 20 and subtract 100.175 squared to obtain variance of 0.656875. Standard deviation is approximately 0.81048.

The updated upper band is approximately 101.79596, the lower band 98.55404 and percentage bandwidth 3.23625%. The close at 103 has exceeded the updated band and widened the distribution, but it has not exceeded the prior structural high of 103.5.

Outside-band closing and rising bandwidth here are two consequences of the same new observation, not independent proof that resistance has been broken. This does not predict a reversal at 103. Price could consolidate and later clear 103.5, or fail and return toward 100. The present evidence is expansion near a structural boundary. The guide to support, resistance and false breakouts explains how to define that separate boundary.

A probe is not automatically a failed trade

Define a false breakout relative to a price boundary and observation period chosen beforehand. An intrabar move above 103.5 followed by a close inside the range is a probe. If the entry rule requires an outside close, it never triggered. Counting it later as a losing trade confuses rejected candidates with positions that actually carried risk.

Once an outside close has triggered the rule, a quick return to the old range and a break of a previously designated pullback low may invalidate it. Conversely, subsequent closes remaining outside, intact higher-low structure and increased activity during comparable trading hours can strengthen evidence of acceptance. Those conditions still need testing; two closes or one volume ratio do not guarantee success.

Confirmation has a price. An executable entry at 104 with invalidation at 102 has a distance of 2. Waiting until entry is 105 while invalidation remains 102 raises that distance to 3. With the same risk budget and other assumptions unchanged, the maximum quantity becomes two-thirds of the earlier amount. Feeling more certain is not a reason to retain the old size.

A filter can improve the win rate and still worsen results if it sacrifices too much reward, adds costs or misses the few large trends. Evaluate the complete outcomes under one executable rule, including rejected and untriggered candidates, rather than collecting attractive breakout screenshots.

Two misleading shortcuts after a squeeze

Touching the upper band is not an automatic short signal. Strong uptrends can remain near or close outside it, just as downtrends can follow the lower band. Fading the first touch after compression may put the position against a newly developing trend.

Returning inside the bands does not independently prove reversal either. After a large rise, the moving average and standard deviation may catch up enough to enclose a price that has barely declined. Check the actual depth of the pullback and whether the breakout boundary or sequence of highs and lows has failed.

Old extreme observations leaving the rolling window can also narrow the bands abruptly even when the newest bars have not become much quieter. Distinguish fresh calm from the expiry of an old shock. Both change the statistic, but they need not carry the same interpretation.

Nor do two standard deviations mean that future prices have a calibrated 95% chance of remaining inside. Price distributions, serial dependence, rolling estimation and changing conditions do not support that simple inference. The bands are not an automatically valid prediction interval.

Data conventions and execution can dominate the pattern

Equity announcement gaps can cross an entire band in one step, and adjustment settings can change historical averages and deviations. A continuous futures chart can jump because of a contract roll. If expansion appears only in a processed series, check the actual contract or tradable price before attributing it to fresh orders.

Spot FX and OTC gold require attention to session and quote source. A move of the same size is not equally executable when spreads and liquidity differ. Exchange futures offer contract-specific volume and open interest, whereas quote-update counts are not total market volume. The volume guide separates those datasets.

A pre-announcement squeeze may reflect participants waiting. The first post-release expansion can continue or reverse after a two-way move. Small historical bandwidth does not ensure a small next shock or execution at a stop's trigger price. If one OHLC bar covers both entry and stop, the bar alone cannot reveal which happened first; use finer data or conservative assumptions.

Keep the evidence that identifies the real problem

A useful record retains the historical bandwidth rank, structural boundary, first qualifying close, executable entry, invalidation and costs. Set parameters before seeing subsequent prices, and keep candidates that stayed narrow, failed immediately or never triggered.

If most mistakes are closes outside the band but below the prior high, the missing distinction is statistical versus structural boundaries. If confirmed breakouts mainly lose during news, execution and event constraints may matter more. A squeeze identifies a volatility condition; whether it is worth trading depends on acceptance of a new range and the cost of taking the risk.