A Bollinger Band squeeze is a stretch of unusually narrow bands, a sign that volatility has dried up, and the textbook claim is that a big move follows. On $SPY, $QQQ, $IWM and $DIA over their full histories, it did not. Across 227 squeezes, the next 20 sessions moved less, on average, than a random day's next 20 sessions in all four ETFs. What the squeeze did deliver was speed: a band broke within 20 sessions 96% of the time, usually within three to five sessions, and 71.1% of those first breaks still pointed the right way at day 20.
So the squeeze is a timing tool, not a size tool. Here is how it was tested, where it helps, and where all four ETFs sit after the November 18, 2026 close.
What Bollinger Bands and BandWidth measure
Bollinger Bands are three lines built from the closing price. The middle band is a 20-session simple moving average. The upper and lower bands sit two standard deviations above and below it. On his own site, John Bollinger writes that the defaults are still 20 periods and two standard deviations, and that the bands use the population calculation for standard deviation. This study does the same.
BandWidth turns the gap between the bands into one number:
BandWidth = (upper band − lower band) ÷ middle band × 100
A worked example from the November 18, 2026 close in $SPY: the upper band was $779.53, the lower band $727.91 and the middle band $753.72. That gives (779.53 − 727.91) ÷ 753.72 × 100 = 51.62 ÷ 753.72 × 100 = 6.85%. Because BandWidth is scaled by price, a reading from 1995 can be compared with one from 2026.
Low BandWidth means the last 20 closes were tightly bunched. StockCharts ChartSchool's Bollinger Band Squeeze page quotes Bollinger's observation that "periods of low volatility are often followed by periods of high volatility," and it looks for BandWidth "near the low end of its six-month range." It also warns that "narrowing bands do not provide any directional clues," and it relays Bollinger's warning about the "head fake," where price breaks one band and then reverses.
How the squeeze was defined for this test
- Data: Yahoo Finance daily bars. $SPY runs from January 29, 1993, $DIA from January 20, 1998, $QQQ from March 10, 1999 and $IWM from May 26, 2000, all through the November 18, 2026 close. Bands use the split-adjusted closing price, not dividend-adjusted prices.
- Squeeze: BandWidth at its lowest reading of the last 126 sessions (about six months), counting the current session.
- Cooldown: after a squeeze is counted, new ones are ignored for 20 sessions, so one long, quiet stretch is not counted a dozen times.
- Size of the move: two measures over the next 20 sessions. One is the absolute % change from the squeeze-day close to the close 20 sessions later. The other is the high-low range, meaning the highest high minus the lowest low over those 20 sessions, divided by the squeeze-day close.
- First break: the first close above that day's upper band or below that day's lower band within 20 sessions. If neither happened, the result is "neither."
- Held: an upper break held if the day-20 close was above the squeeze-day close. A lower break held if it was below.
- Baseline: the same measures for every day with a valid band and a full 20-session window. That is 29,236 days across the four ETFs.
The rules produced 59 squeezes in $SPY, 61 in $QQQ, 54 in $IWM and 53 in $DIA. Every one has a complete 20-session outcome. The latest ones date from July to September 2026.
Did a bigger move follow the squeeze?
| ETF | Squeezes | Avg absolute 20-session change: squeeze / all days | Median absolute change: squeeze / all days | Avg 20-session high-low range: squeeze / all days | BandWidth 20 sessions later vs squeeze day (median) |
|---|---|---|---|---|---|
| $SPY | 59 | 2.90% / 3.35% | 2.23% / 2.68% | 5.76% / 6.69% | 1.49x |
| $QQQ | 61 | 4.20% / 4.99% | 3.68% / 3.92% | 8.73% / 9.74% | 1.67x |
| $IWM | 54 | 4.03% / 4.56% | 2.74% / 3.68% | 8.23% / 9.08% | 1.83x |
| $DIA | 53 | 2.61% / 3.34% | 2.01% / 2.62% | 5.50% / 6.62% | 1.70x |
Price close, full history through November 18, 2026. "All days" includes every session with a full forward window, so those windows overlap.
On both measures and in every ETF, the 20 sessions after a squeeze moved less than the average 20-session stretch. The average absolute change ran about 12% to 22% below the all-day figure. Tightening the definition to a 252-session low did not reverse it in any of the four: $SPY's 30 squeezes on that rule averaged 2.47% against 3.35% for all days.
The last column explains the apparent contradiction. Volatility did expand after a squeeze. Median BandWidth was 1.49 to 1.83 times wider 20 sessions later, while on a typical day it barely changed (0.98x to 1.01x). Relative to its own narrow band, the next 20-session range was large, 1.6 to 1.9 band-widths against about 1.06 to 1.07 for a typical day. But expansion started from a very low base. The median $SPY squeeze began at a BandWidth of just 2.96%. Low volatility tends to cluster, and a 50% to 80% widening of a small number is still a small number.
Which band broke first, and did the break hold?
| ETF | Upper first | Lower first | Neither in 20 | Upper break held | Lower break held | Any first break held (all days) |
|---|---|---|---|---|---|---|
| $SPY | 32 (54.2%) | 24 (40.7%) | 3 (5.1%) | 25 of 32 (78.1%) | 15 of 24 (62.5%) | 71.4% (78.3%) |
| $QQQ | 33 (54.1%) | 26 (42.6%) | 2 (3.3%) | 28 of 33 (84.8%) | 18 of 26 (69.2%) | 78.0% (79.5%) |
| $IWM | 26 (48.1%) | 28 (51.9%) | 0 (0.0%) | 18 of 26 (69.2%) | 21 of 28 (75.0%) | 72.2% (77.2%) |
| $DIA | 28 (52.8%) | 21 (39.6%) | 4 (7.5%) | 18 of 28 (64.3%) | 12 of 21 (57.1%) | 61.2% (78.0%) |
| All four | 119 (52.4%) | 99 (43.6%) | 9 (4.0%) | 89 of 119 (74.8%) | 66 of 99 (66.7%) | 71.1% (78.3%) |
Three things stand out.
- A break came fast and almost always. Only 9 of 227 squeezes went 20 sessions without a close outside the bands. On all days, 20.5% did. The median first break came three to five sessions after the squeeze, against seven to eight sessions on all days. Narrow bands are easy to close outside of.
- The direction was close to a coin flip. The upper band broke first 52.4% of the time. On all days the upper band also broke first more often (41.5% versus 38.1%), so the lean reflects a rising market more than the squeeze. $IWM broke lower more often than higher.
- The first break was a decent but imperfect tell. It held at day 20 in 71.1% of squeezes, below the 78.3% rate for first breaks on ordinary days. The gap was widest in $DIA (61.2% versus 78.0%). Roughly three in ten squeeze breaks were head fakes on this definition.
Recent examples show both outcomes. $IWM squeezed on February 20, 2026, broke its lower band on March 5, and was 8.46% lower 20 sessions later. It squeezed again on July 13, broke lower on July 29, and was 2.21% higher by day 20: a head fake.
Where BandWidth stands at the November 18, 2026 close
| ETF | Close | Lower / upper band | BandWidth | Percentile, full history | Last squeeze |
|---|---|---|---|---|---|
| $SPY | $730.20 | $727.91 / $779.53 | 6.85% | 68th | Aug 28, 2026 |
| $QQQ | $659.51 | $641.82 / $730.78 | 12.96% | 81st | Sep 14, 2026 |
| $IWM | $270.12 | $268.97 / $286.64 | 6.36% | 38th | Jul 13, 2026 |
| $DIA | $501.73 | $505.95 / $533.64 | 5.33% | 50th | Sep 3, 2026 |
The percentile is the share of all daily BandWidth readings since each ETF's launch at or below the November 18 reading.
No ETF is in a squeeze now. All four squeezed between July and September 2026, and the bands have since widened. $QQQ's BandWidth of 12.96% is five times its 126-session low of 2.58% from September 15, and is higher than about 81% of its history. $SPY closed just 0.31% above its lower band, while $DIA closed 0.83% below its lower band, which is the expansion phase rather than the quiet one. Source: Yahoo Finance historical data; the readings move with every close.
Because the squeeze says little about direction, traders often pair it with a separate breakout rule, such as a Donchian channel breakout, and with a volatility-based exit, such as an ATR trailing stop, to cap the cost of a head fake.
Limits: one data source, closes rather than intraday fills, no costs, taxes or slippage, overlapping baseline windows, and four ETFs that all survived the period. More explainers sit in Analysis and on the Traders On Trend home page.
Bull/Bear Verdict
Bull Case: Where it helps: a squeeze reliably flagged that volatility was about to expand from its own low. BandWidth was about 1.5 to 1.8 times wider 20 sessions later, a band broke in 96% of cases, and the first break held at day 20 about 71% of the time.
Bear Case: Where it misleads: the "big move" was smaller than an ordinary 20-session move in all four ETFs, the break direction was close to 50/50, and the first break held less often than on ordinary days. About three squeeze breaks in ten were head fakes.
FAQ
What is a Bollinger Band squeeze?
It is a period when the 20-session, two-standard-deviation Bollinger Bands are unusually narrow, measured by BandWidth. This test counts a squeeze when BandWidth hits its lowest level in 126 sessions, with a 20-session cooldown between signals.
Does a Bollinger Band squeeze predict direction?
Not on this data. Across 227 squeezes in $SPY, $QQQ, $IWM and $DIA, the upper band broke first 52.4% of the time and the lower band 43.6% of the time. Traders use the first band break, not the squeeze itself, to pick a side.
How big is the move after a squeeze?
Historically, smaller than average in absolute terms. The average absolute 20-session change after a squeeze was 2.90% in $SPY, against 3.35% for all days. Relative to the narrow bands, though, the move was large, and BandWidth typically expanded by half or more.
What BandWidth counts as low?
It depends on the ETF. The median squeeze began at a BandWidth of 2.96% in $SPY, 3.88% in $QQQ, 3.81% in $IWM and 2.57% in $DIA. Comparing a reading with the ETF's own six-month range matters more than any fixed number.