How to backtest a Bollinger Bands strategy on crypto

Bollinger Bands get used two opposite ways, and confusing them is why so many band strategies fail. The bands are a moving average plus and minus a multiple of standard deviation, so they measure volatility — they don't predict direction. You can fade a touch of the band (mean reversion) *or* trade a breakout out of a squeeze (volatility expansion), but those are different strategies for different regimes. Backtesting is how you find out which one actually works on your market, and when.

Step 1: decide which Bollinger strategy you're testing

First, pick a side. Mean reversion assumes price snaps back to the middle band: enter when price closes outside a band, exit at the midline. It works in ranges and gets destroyed in trends, where price "walks the band" for weeks. Squeeze breakout assumes a period of low volatility (narrow bands) precedes a big move: enter when the bands expand and price breaks the band, ride the expansion.

Whichever you choose, you still need all four decisions: entry trigger, exit, stop-loss, and a regime filter so you're only trading the conditions your variant is built for. A mean-reversion fade needs a "ranging" filter; a squeeze breakout needs confirmation that the move is real, not a fakeout.

Step 2: get clean historical data

Your backtest is only as honest as its data. Use real OHLCV candles for the exact symbol, timeframe, and exchange you'll trade, over a window covering more than one regime. This matters doubly for Bollinger: a mean-reversion fade will look brilliant on a ranging year and catastrophic on a trending one, so a single-regime test tells you nothing about live behaviour.

Step 3: run it and read the right metrics

Total return hides the risk you took. Read maximum drawdown, the Sharpe ratio, profit factor, and win rate with average win/loss. Mean-reversion band fades typically show high win rates with the occasional brutal loss when price walks the band — exactly the profile that blows up an account if the stop is loose.

Here is a complete squeeze-breakout version — it waits for low volatility, then trades the expansion with a trend confirmation:

strategy:
  name: bollinger_squeeze
  indicators:
    - { id: bb, kind: Bollinger, period: 20, mult: 2.0 }
    - { id: adx, kind: ADX, period: 14 }
  rules:
    entry:
      all:
        - { type: above, left: adx, right: 20 }        # a real move, not chop
        - { type: crosses_above, left: close, right: bb.upper }
    exit:
      any:
        - { type: crosses_below, left: close, right: bb.middle }
  risk:
    size_pct: 0.5
    stop_loss_atr: 2.0

Step 4: the test that actually matters — walk-forward

Bollinger Bands have two tempting knobs — the period and the standard-deviation multiplier — and tuning both against your whole history is the fastest way to curve-fit a beautiful, useless equity curve. Walk-forward optimization tunes on one window and tests on the *next* window the optimizer never saw, then rolls forward. If your 20-period, 2.0-sigma settings only worked because you hand-picked them on the full dataset, walk-forward will show the edge collapsing on unseen data.

This is where the overfitting score earns its place. A Bollinger strategy that survives walk-forward across a rally, a crash, and a chop phase is worth paper-trading; one that only shines in-sample is a memory of the past.

Do it in Noon Barbari

You can build both Bollinger variants without code. Add Bollinger Bands and ADX from the indicator library, wire the entry/exit/stop in the visual designer, and run the backtest over years of crypto data in seconds. Then run walk-forward to see whether your period and multiplier are real parameters or noise. The free tier covers one full strategy end to end — enough to test this exact rule and read the robustness score yourself.