A band touch is guaranteed to happen
Bollinger Bands are a moving average with an upper and lower band placed a number of standard deviations away, measured over the same window. That construction has a consequence people build strategies without noticing:
A band touch is not a rare event. Its frequency is set by the indicator's own formula.
If the bands sit two standard deviations from a 20-period average, then by construction a predictable minority of the last 20 closes will fall outside them. Price reaching the band is the indicator working normally, not the market doing something unusual.
So "price touched the upper band" carries almost no information on its own. What does carry information is how wide the bands are, because that is a direct readout of recent volatility.
The timing problem nobody mentions
The bands are computed from the same window the volatility has just entered. So they widen after a move begins, not before.
That produces a specific and awkward sequence:
- the market is quiet, so the bands are narrow
- a real move starts
- price is immediately far outside a band that is still narrow from step 1
- the bands widen over the following bars, catching up
A band-touch rule fires hardest at step 3 — which is either the best moment (the move is genuine and you are early) or the worst (the move is noise and the band was simply too tight). The indicator cannot distinguish those, because both look identical at the moment of the touch.
Which is why breakout and reversion are opposite bets on the same event
| You believe | You trade | You lose when |
|---|---|---|
| the move is real | a breakout: enter in the direction of the touch | the move is noise |
| the band was too tight | reversion: enter against the touch | the move is real |
Both strategies use the same trigger. They are not variations of one idea — they are opposite predictions, and which works depends entirely on the regime. That is why a band strategy needs a regime filter more than it needs better band parameters.
Building the breakout version
| Field | Value | Note |
|---|---|---|
| Bollinger length | 20 | conventional |
| Deviations | 2 | conventional; changing it changes the touch frequency directly |
| Entry | close above the upper band | |
| Regime filter | ATR above its own average | only take breakouts when volatility is expanding |
| Exit | return to the middle band, or an ATR stop |
The regime filter is the part that matters, and the construction — a dimensionless ATR ratio rather than a fixed ATR level — is in the ATR regime filter.
Two things to check in the export
The deviation multiplier must be an input you can move. Changing it from 2 to 2.5 should visibly change the trade count. If it does not, the multiplier is compiled in as a literal and the control is inert.
The middle band is a moving average, so if you also add a separate moving-average trend filter with the same period, you have added the same series twice — the double-count problem described in the MACD playbook.
Tactix AI — Studio vs Guide
Tactix AI is AlfaTactix’s product assistant brand (open Tactix AI).
- Tactix Studio turns a one-sentence strategy description into a draft across Timeframe, Signals, Filters, and Risk in the visual Strategy Builder. You review and edit every field before Code Generator writes MQL5 or Pine Script.
- Tactix Guide explains the step you are on — what to fill, what a control means, or how to phrase a rule — without dumping untested source code.
That is form-first automation: the LLM never replaces Code Generator, and you keep plan limits and real-time validation.