Originally Published: · Last Updated:
Does the Bollinger Squeeze work? It identifies unusually low volatility, but it is not a complete strategy and has no universal win rate. The 7,846-rule studies did not directly test it, so any squeeze strategy still needs a defined entry, exit, cost model, and out-of-sample test.
What the Squeeze Actually Tells You
A Bollinger Squeeze occurs when BandWidth contracts to a relatively low level. It tells you that recent price dispersion has narrowed. That can be useful because quiet periods sometimes precede larger moves, but the setup does not specify the direction, timing, holding period, or exit.
That distinction is not a criticism added by skeptics. On his official site, John Bollinger describes a squeeze as a potential indication of a new trend and says it is not a signal by itself. He points traders back to price action and supporting indicators as the final arbiters. His published rules also warn that a touch of either band is not, by itself, a buy or sell signal.
Primary source: Bollinger Band Squeeze guidance and John Bollinger’s official rules.
The setup answers: Is volatility unusually compressed relative to the recent past?
It does not answer: Direction, entry, exit, or position size.
This is why two traders can both claim to trade “the Bollinger Squeeze” while using materially different systems. One buys an upside close outside the upper band. Another trades either breakout direction. A third combines BandWidth with volume, trend, or momentum. A fourth buys an options straddle before an event. Those systems do not share the same payoff distribution, turnover, or cost structure.
Why There Is No Single Bollinger Squeeze Win Rate
A win rate belongs to a fully specified rule, not to an indicator name. Change the asset, threshold, entry timing, stop, profit target, holding period, or trade direction and the win rate changes. Even the definition of a squeeze varies: a six-month BandWidth low, the lowest decile of BandWidth, Bollinger Bands moving inside Keltner Channels, or a discretionary visual contraction.
| Rule component | Examples that change the result |
|---|---|
| Squeeze definition | Six-month BandWidth low, rolling percentile, or Bollinger-inside-Keltner condition |
| Direction | Long only, short only, breakout direction, or market-neutral |
| Entry | Close outside a band, next-day open, intraday stop order, or confirmation bar |
| Exit | Middle-band cross, fixed target, trailing stop, time stop, or opposite signal |
| Universe | SPY, individual large-cap stocks, small caps, futures, crypto, or options |
| Costs | Spread, slippage, commissions, borrow fees, option premium decay, and taxes |
A vendor can report a high hit rate by using a small sample, a favorable market, a loose definition of success, or a profit target much smaller than the stop. A lower win-rate system can still make money if average winners are larger than average losers. A high win-rate system can lose money for the opposite reason. The useful outputs are net return, drawdown, turnover, exposure, average win, average loss, and performance on unseen data, not win rate alone.
What the 7,846-Rule Studies Prove
Sullivan, Timmermann, and White assembled 7,846 technical trading rules to study data snooping. Bajgrowicz and Scaillet later re-examined the same universe on daily Dow Jones Industrial Average data from 1897 through July 2011 using a False Discovery Rate framework, persistence tests, and transaction costs.
The rule universe contained five categories: filter rules, moving averages, support and resistance rules, channel breakouts, and on-balance-volume averages. In the papers, a channel breakout is based on price moving through parallel trend lines. A Bollinger Squeeze is based on contraction in a volatility band built from a moving average and rolling standard deviation. Similarity at a high level does not make them the same tested rule.
That means the studies cannot support claims such as “the Bollinger Squeeze was one of the 7,846 rules,” “its directional win rate was about 50%,” or “the papers proved the squeeze loses after costs.” Those statements transfer a result from one rule family to a different, untested setup.
What the studies do support is still important. Bajgrowicz and Scaillet found that investors could not reliably select future winning rules from past winners, and low transaction costs offset the apparent economic value of many rules. In the three most recent Dow subperiods in their study, the authors detected no positive performance among the tested rules even before transaction costs. That is evidence against casual trust in backtested technical rules, not a direct Bollinger Squeeze verdict.
Primary sources: Sullivan, Timmermann, and White (1999) · Bajgrowicz and Scaillet (2012).
Hsu and Kuan reached a related but more qualified result. Their data-snooping-corrected study found profitable technical rules in the relatively younger NASDAQ Composite and Russell 2000 samples, but not in the mature Dow and S&P 500 samples. Again, that finding concerns the rules they tested. It is not proof that a Bollinger Squeeze works in small caps.
Primary source: Hsu and Kuan (2005).
How to Turn the Setup Into a Testable Rule
The phrase “trade the squeeze” is too vague to test. Before opening a backtest, write the rule in a form that another person could implement without seeing your chart.
| Field | What to record before testing |
|---|---|
| Asset universe | Name the index, ETF, futures contract, or stock-selection rule. Include delisted securities when testing stocks. |
| Squeeze condition | Define the exact BandWidth formula, lookback, and threshold. |
| Direction rule | State whether the strategy goes long, short, both, or remains direction-neutral. |
| Entry | Specify the signal bar, execution bar, price assumption, and order type. |
| Exit | Specify stops, targets, time limits, and what happens after a head fake. |
| Position sizing | Define account risk, leverage, and whether positions overlap. |
| Benchmark | Choose buy-and-hold, cash, volatility exposure, or another risk-matched alternative. |
| Costs | Model commissions, spread, slippage, borrow, financing, and option-specific costs where relevant. |
| Validation | Freeze the rule before opening the holdout period. Report every parameter variant searched. |
A backtest is not ready merely because code runs. The rule should use a one-bar lag when the signal depends on the closing price, unless the strategy can genuinely execute at that close. Costs should be charged when positions change, not added later as a generic annual haircut. Options tests need implied-volatility data and realistic bid-ask assumptions; an equity-price backtest cannot prove that buying straddles around squeezes is profitable.
The related guide on how to audit a technical-analysis backtest walks through search bias, implementation costs, and unseen-data testing. The related TradingView settings guide explains why spread and slippage should match the instrument and execution method rather than come from a universal percentage.
Status: No universal squeeze backtest is reported here because “Bollinger Squeeze” is a setup, not a complete rule.
What would change that: A pre-specified definition covering the asset universe, squeeze threshold, direction, entry, exit, sizing, benchmark, costs, and holdout period.
Evidence ceiling: The academic 7,846-rule literature is used to set the testing standard. It is not used as direct performance evidence for the Bollinger Squeeze.
Decision Router: Reject, Paper-Test, or Allocate
| Evidence state | Action | Reason |
|---|---|---|
| Only a chart screenshot or claimed win rate | Reject the claim | The rule and measurement cannot be reproduced. |
| Rule is defined, but the search history or costs are hidden | Paper-test only | The attractive result may reflect tuning or unrealistic execution. |
| Rule is frozen and costs are documented, but no untouched sample exists | Forward-test without capital | Live observation can reveal fills and rule ambiguity without financial risk. |
| Rule survives multiple unseen periods and realistic costs | Consider a limited allocation | The evidence is stronger, but regime and model risk remain. |
| Options strategy relies on volatility expansion | Require options-specific evidence | Direction-neutral does not mean profitable; implied volatility, decay, and spread determine the result. |
For most retail traders, the productive next step is not adding another confirming indicator. It is writing down the existing rule and running it unchanged in a paper account. Record the signal time, intended entry, actual fill, exit reason, spread, slippage, and whether a discretionary exception was made. A rule that changes after every loss cannot generate honest forward evidence.
Bollinger Squeeze FAQ
Does a Bollinger Squeeze predict direction?
Not by itself. It identifies volatility contraction. John Bollinger’s official guidance says a squeeze is not a signal on its own and that price action plus supporting indicators determine what to do.
What is the Bollinger Band Squeeze win rate?
There is no universal figure. The answer depends on the exact squeeze definition, asset, direction rule, entry, exit, costs, and sample. A percentage without those fields is not reproducible.
Did academics test the Bollinger Squeeze among 7,846 rules?
The cited 7,846-rule universe included filter rules, moving averages, support and resistance rules, price-channel breakouts, and on-balance-volume averages. The papers do not list the Bollinger Squeeze as a tested rule. Their findings should not be presented as a direct squeeze backtest.
Can a direction-neutral options trade solve the problem?
It removes the need to predict up versus down, but it does not guarantee a profit. A long straddle or strangle must overcome the premium paid, time decay, changes in implied volatility, and the bid-ask spread. It needs its own options-specific backtest.
Is a BandWidth percentile filter enough?
No. A percentile can make the squeeze definition reproducible, but the strategy still needs a direction rule, entry, exit, position size, benchmark, costs, and unseen test period.
The Bottom Line
The Bollinger Squeeze is useful as a description of volatility contraction. It is not a complete trade and it does not come with a defensible universal win rate.
Before risking money, write down the exact rule. Then ask three questions: Was it selected from many variants? Does it survive realistic costs? Does the frozen version work on data it never saw? Until those questions have evidence, treat the squeeze as a setup to investigate, not an edge you own.
Your turn: Could another trader reproduce your squeeze rule from one written paragraph, without looking at your chart?
- John Bollinger: Official Bollinger Bands rules and Squeeze guidance were used to define what the setup does and does not claim.
- Sullivan, Timmermann, and White (1999): Used for the data-snooping framework and the composition of the 7,846-rule universe.
- Bajgrowicz and Scaillet (2012): Used for the rule-universe descriptions, persistence findings, modern Dow subperiod results, and transaction-cost methodology.
- Hsu and Kuan (2005): Used to show that results varied across mature and younger equity-index samples, without transferring those results to the squeeze.
Method: This revision separates direct evidence about the Bollinger Squeeze from broader evidence about technical-rule testing. No win rate or dollar-loss estimate is reported because the sources do not provide one for a fully specified squeeze strategy.
Editorial transparency: AI tools assisted with drafting and consistency checks. Danny Hwang reviewed the final article, source scope, and calculations. No source was treated as direct evidence for a rule it did not test.
Disclosure: TheFinSense does not sell trading signals or receive compensation from the indicator, broker, or platform providers discussed in this article.
Educational quantitative analysis based on published data. Not investment, tax, or legal advice. Consult a licensed professional before acting on any calculation. About TheFinSense.