Knowledge · Strategies · Bollinger Band Mean Reversion

Bollinger Band Mean Reversion

John Bollinger (1980s); mean-reversion variant popularized by various authors
Mean Reversion Evidence: Moderate swing equitiesfxcryptocommodities
6/10
Relevance for Botty
Long when price closes below the lower band + returns back inside; exit at the middle band (20-SMA). Volatility-scaled.
Core concept
Bollinger Bands (20-SMA ± 2 standard deviations) define a statistically 'normal' range. Price outside the bands is considered extreme — in ranges, mean reversion toward the mean follows. The system works regime-dependently: in bear markets it does better as a breakout, in ranges it does better as mean reversion.
Relevance for Botty
Relevance Score 6/10
BTC perp on 4h or 1h — Bollinger mean reversion works in consolidating phases, fails in strong trends. Only sensible as a regime-conditional add-on: e.g. active when ADX < 20 (no trend). Dangerous as a stand-alone strategy without a regime filter.
Rules

Entry

  • Long entry: close below the lower band (lower = SMA-20 - 2×stdev), then close back inside the bands
  • Short entry: analogously at the upper band
  • Optional filter: RSI < 30 for additional confirmation

Exit

  • Target 1: middle band (20-SMA) — higher win rate
  • Target 2: opposite band — higher R:R
  • Stop: fixed value below the pullback low, or 2×ATR
Parameter
NameTyp. valueDescription
period 20 SMA for the middle band
stdev_multiplier 2.0 Width of the bands; 2.5 is often better for crypto
rsi_filter < 30 for long Optional, for confirmation
Pros & Cons

Pros

  • Versatile — mean reversion OR breakout depending on regime
  • Self-scaling with volatility (bands expand/contract)
  • Visually intuitive, easy to backtest
  • A good building block for regime-adaptive systems

Cons

  • In strong trends mean reversion fails completely — price clings to the band
  • Unreliable without a regime filter
  • Crypto frequently overshoots the classic 2-stdev bounds
  • Optimal parameters are highly regime-dependent (no set-and-forget)
Typical performance
notes
The middle band as target delivers a higher win rate; the opposite band gives more R:R but a lower win rate. Strong regime dependence.
win rate
~44-60% depending on market and exit strategy
max drawdown
~4-15%
profit factor
~1.4-1.8
Bot suitability
Good for medium-frequency bots; requires regime detection for stable performance.
Background

Core idea

Bollinger Bands are a volatility envelope: an SMA-20 in the middle, ±2 standard deviations (based on the last 20 closes) as the outer bands. Statistically, ~95% of prices should lie within the bands (under a normal-distribution assumption — in reality fewer, due to fat tails).

Two opposing uses:

  1. Mean reversion: price outside the bands is 'extreme' → a return toward the mean is likely.
  2. Breakout: a strong move through the band signals the start of a trend → follow-through is likely.

Which one works depends on the market regime — and this is exactly where naive Bollinger systems are weak.

The mean-reversion rule

Close < LowerBand  →  watch
next candle: Close > LowerBand  →  Long
Exit: price reaches MiddleBand (SMA-20)

The trick is waiting for the return: if price merely punches through the band and keeps falling, nothing is bought — only when it stabilizes.

Regime dependence

An academic paper on BTC/USDT (Arda 2024) shows clearly:

  • Bull regime: mean reversion profitable, breakout below average
  • Bear regime: mean reversion loses, breakout wins
  • Ranging: mean reversion clearly superior

Without regime detection the system is unpredictable. The band squeeze (contracting bands, low vol) followed by a breakout is a separate strategy of its own, not covered here.

Typical parameters for crypto

  • 20/2.0 is standard, but on 1h BTC the 2-stdev bounds overshoot too often
  • Recommended: 20/2.5 or 50/2.0 for more stable signals
  • Combine with an ADX < 20 filter for range confirmation

Relevance for Botty

Botty's current strategies are trend-/breakout-oriented. Bollinger mean reversion would be a complementary range module that is active only when ADX is below a threshold (market consolidating). Stand-alone it is too risky, because BTC 'rides the band' more often than equity indices do.