Bollinger Bands Mean Reversion EA MQL5 Code & Strategy

Why Most Bollinger Band EAs Fail (And How to Fix It)

I've lost count of how many Bollinger Band EAs I've seen that simply buy when price touches the lower band and sell when it hits the upper band. They work great in backtests on trending data—until you run them forward on a ranging market. Then they blow up. The problem isn't the indicator; it's the assumption that touching a band guarantees a reversal.

Mean reversion with Bollinger Bands isn't about catching every band touch. It's about identifying when price has overshot its statistical range during a period of low volatility, then fading that move with a tight risk model. In this post, I'll walk you through building a real MQL5 Expert Advisor that does exactly that—complete with entry filters, a trailing stop that adapts to volatility, and position sizing that won't wipe you out on the third trade.

This isn't a copy-paste script. I'll explain why each piece matters, show you the code, and point out the pitfalls that backtests hide. If you're writing your first mean reversion EA or improving an existing one, you'll find something useful here.

The Strategy Logic: Beyond Band Touches

Standard Bollinger Bands use a 20-period SMA with two standard deviations. In a normal distribution, about 95% of price action falls within the bands. When price pierces the outer band, it's statistically extreme—but that doesn't mean it will reverse immediately. Strong trends can ride the band for dozens of bars, as anyone who's shorted a breakout on the upper band knows too well.

Our EA adds two critical filters that separate decent entries from disaster:

• RSI confirmation: We only enter a mean reversion trade if RSI(14) is below 30 (for a long) or above 70 (for a short). This prevents fading a move that still has momentum. I've seen plenty of traders buy the lower band while RSI is at 45—that's not oversold, that's a knife waiting to catch your face.
• Band width filter: We only trade when the Bollinger Band width (upper minus lower, divided by middle) is below a threshold. Narrow bands mean low volatility, which historically precedes explosive moves—but for mean reversion, we want to catch the snap-back before the explosion. If bands are wide, volatility is high and reversals are less reliable.

The entry itself is simple: place a pending order at the upper or lower band, but only when RSI confirms overextension and the bands are narrow enough. If price doesn't reverse within a few bars, the order expires. This avoids catching falling knives—a mistake that's cost me more than I care to admit in early versions of this EA.

Why the Band Width Filter Matters More Than You Think

Let me be blunt: most traders ignore band width entirely. They see a touch of the lower band and jump in, regardless of whether volatility is contracting or expanding. Here's the reality check I learned from a nasty drawdown on EURUSD in 2022. When bands are wide—say, width above 0.10 on a 20-period setting—price often keeps pushing through the band because volatility is accelerating. That's a trend, not a reversion setup. Narrow bands, typically below 0.05 on daily charts, indicate a compression phase. That's when mean reversion has a higher probability of working, because price tends to snap back to the mean before the next expansion.

I've tested this on GBPUSD and USDJPY across H1 and H4 timeframes. The band width filter alone improved win rate by about 12% in my forward tests, compared to a version that only used RSI. It's not a magic bullet—nothing is—but it's a solid edge that most public EAs miss.

MQL5 Implementation: The Core Code

Let's get into the MQL5 code. I'll assume you know the basics of creating an Expert Advisor in MetaEditor. If you're new, create a new EA from the wizard, select "Expert Advisor (template)", and we'll replace the boilerplate. Open MetaEditor from the Tools menu in MT5, then File -> New -> Expert Advisor. Name it something descriptive like "BB_MeanReversion_EA".

Input Parameters

Set these at the top of your EA file. Every parameter should have a sensible default and a clear comment—future you will thank yourself when you're debugging at 2 AM before a London open. I use Hungarian notation for MQL5 because it keeps variable types obvious when you're scanning code quickly.

Indicator Handles and Initialization

In MQL5, we use indicator handles instead of direct calls like in MQL4. In OnInit(), create handles for Bollinger Bands, RSI, and ATR. Here's the setup:

int bbHandle, rsiHandle, atrHandle;

int OnInit() {
    bbHandle = iBands(_Symbol, _Period, BB_Period, 0, BB_Deviation, PRICE_CLOSE);
    rsiHandle = iRSI(_Symbol, _Period, RSI_Period, PRICE_CLOSE);
    atrHandle = iATR(_Symbol, _Period, 14);
    if(bbHandle == INVALID_HANDLE || rsiHandle == INVALID_HANDLE || atrHandle == INVALID_HANDLE) {
        Print("Failed to create indicator handles. Error: ", GetLastError());
        return INIT_FAILED;
    }
    return INIT_SUCCEEDED;
}

Notice I use iBands with the shift parameter set to 0—meaning we use the current bar's close. Some developers prefer shift=1 to avoid repainting on the current bar. I'll address that in the risks section, but for now, this works for bar-close checks. Also, I hardcoded ATR period to 14 because it's standard, but you could make it an input parameter if you want to optimize across timeframes.

Entry Logic in OnTick()

Here's where the strategy comes together. We check for new bars first to avoid processing the same tick multiple times. Then we copy indicator buffers and evaluate conditions.

void OnTick() {
    static datetime lastBarTime = 0;
    if(Time[0] == lastBarTime) return;
    lastBarTime = Time[0];

    double bbUpper[3], bbMiddle[3], bbLower[3];
    double rsiVal[3];
    double atrVal[3];

    if(CopyBuffer(bbHandle, 0, 0, 3, bbMiddle) < 3) return;
    if(CopyBuffer(bbHandle, 1, 0, 3, bbUpper) < 3) return;
    if(CopyBuffer(bbHandle, 2, 0,