Bollinger Squeeze Breakout
BullishContinuationIntermediateBands squeeze tight, then a close above the upper band. The squeeze added nothing.
Verdict: SkipBitcoin 4H: 332 trades, won 35%, avg trade +0.2% vs random day +0.5%. Backtest
Measure how wide the Bollinger bands are compared with the average. A squeeze is when that width is in the narrowest fifth of the last 120 candles.
step 1/4
Trade rules
| Direction | Long (buy) |
|---|---|
| Entry | The Bollinger bandwidth ((upper - lower) / middle band, BB(20, 2)) was at or below its 20% quantile over the last 120 candles on at least one of the 5 candles before, and a candle CLOSES above the upper band after closing at or below it the candle before. Buy at that close; exit at the first close below the middle band. Paper traded on 1-hour and 4-hour candles with that exit and a protective stop inside the candle at the middle band or 2 x ATR14 below the entry, whichever is lower; the encyclopedia backtest uses that stop, a target of two times the risk and a 100-candle limit. Tested in docs/research/2026-10-indicator-tests.md. |
| Cancelled if | No squeeze in the 5 candles before the breakout: then it is a plain Bollinger breakout. A wick through the band that closes inside is not a signal. |
| Target | Two times the risk, projected from the breakout |
4 common mistakes
- Believing the squeeze predicts the big move. In our test it did not: the plain breakout did as well against random entries on every timeframe.
- Judging a squeeze by eye. The tested rule compares the band width with its own last 120 candles.
- Waiting for a squeeze and missing the trades. The filter removed 40-60% of the signals and left the return per year lower.
- Guessing the direction before the break. This rule only buys a close above the upper band.
Backtest
same rules every time · equal size · no fees · method| Market | Trades | Won | Avg trade | Random day |
|---|---|---|---|---|
| Bitcoin · 4H · 2017-2026 | 332 | 35% | +0.18% | +0.46% |
| US large caps (50) · 1D · 2016-2026 | 2060 | 42% | +1.26% | +1.47% |
Verdict: SkipIt made money, but less than buying on a random day.
pattern +61%random +153%
pattern +2593%random +3025%