Bollinger Bands and MACD Strategy — What Six Years of GBP/USD Data Say

By Uladzimir Leuchankou, founder of Signal2Forex

Is a Bollinger Bands and MACD strategy just two indicators on one chart? I suppose that is an answer too. Most pages describe it that way — the bands show where price is stretched, the MACD confirms the direction, and you trade when both agree. But for me a strategy is a set of rules I can write into an expert advisor and run over old data.

So this time I did exactly that. I took 1,458 daily closes of GBP/USD from January 2021 to October 2026, built the bands and the MACD the way MetaTrader draws them, and counted the trades. The numbers below are mine, with pictures, and some of them do not agree with the usual story.

Daily noon rates from the Federal Reserve H.10 release, closes only, no spread and no swap. One pip on GBP/USD is $10 on one standard lot, $0.10 on 0.01 lot.

Two tools, one set of closes

What do the bands measure? Twenty closes, their average and how far they scatter around it. The middle band is a 20-period simple moving average. The upper and lower bands sit two standard deviations above and below it. When the market is quiet, the bands come close together — traders call this a squeeze. When the market moves, they open.

John Bollinger published the bands in 1983 and still keeps his rules on his own site. One of them I repeat to myself before every test: tags of the bands are tags, not signals.

The MACD is older. Gerald Appel built it in the 1970s from two exponential moving averages — a 12-period and a 26-period one. Between them runs the MACD line, their gap. A 9-period average of that gap is the signal line. The histogram is the gap between the two lines. On 30 October 2026 the GBP/USD daily MACD line stood at +0.0007, the signal at −0.0009, so the histogram was +0.0016 — positive, and rising for three weeks.

This is the part I want you to notice. Both tools are built from the same column of closes. The bands take an average and a scatter, the MACD takes two averages and a difference. Bollinger writes that a confirming indicator should not be collinear with the bands — not built from the same arithmetic.

The MACD is close to that line, and I am not sure he would call it a clean second opinion. That doubt is what made me run the numbers.

One more thing on the arithmetic, because people write to me asking for a worked example. On 30 October 2026 the 20-day average of GBP/USD closes was 1.3298 and the standard deviation was 0.0071. Upper band — 1.3298 plus two times 0.0071 — 1.3441. Lower band — 1.3155. Width — 2.15% of the middle. The close that day was 1.3398, inside the bands and above the middle.

If you can do this once in a spreadsheet, you understand the indicator better than most of its users. People search for this pairing as MACD Bollinger Bands, or as a Bollinger Bands MACD strategy, or the other way round as a MACD and Bollinger Bands setup. It is the same two tools on one chart, and the order of the words changes nothing.

The October squeeze on GBP/USD

Here is a live case, and it is the reason I chose this pair. On 1 October 2026 the GBP/USD bands were 3.80% wide. By 19 October the width had fallen to 1.52% — the narrowest since mid-September. The pound had been sliding since the Bank of England held rates on 17 September, and the 14 October close of 1.3154 sat 30 pips above the lower band. Not outside it. Just near.

Then the direction changed. The MACD histogram had already turned positive on 9 October, while price was still falling — the fall was slowing before it ended. On 15 October the close jumped back above the middle band. On 20 October it closed at 1.3376, above the upper band.

From there the pair walked the band for two weeks: 1.3390, 1.3382, 1.3374, a dip to 1.3305, then 1.3398 on 30 October. The MACD line crossed above zero on 29 October. From the 14 October close to 30 October that is 244 pips — $2,440 on one standard lot, $24.40 on 0.01 lot, about 24 cents on 0.01 lot of a cent account.

The move ran into the Bank of England meeting. On 5 November 2026 at 12:00 GMT the Bank raised Bank Rate from 3.75% to 4%, the first rise since August 2023, by eight votes to one — Catherine Mann wanted 4.25%. The pair had reached 1.3489 that morning and barely moved on the news. The rise had been priced in for two weeks, and the bands had already shown the whole move.

So did my rules catch it? No. My re-entry rule had bought on 21 September at 1.3371, held for 15 bars and sold on 13 October at 1.3256 — minus 115 pips, one day before the low. Since the 14 October close never went outside the lower band, there was no new signal.

The breakout rule wanted a MACD cross within five bars of the break, and the cross was eleven bars old by 20 October. The best move of the autumn, and both rules watched it from the side.

Bollinger Bands and MACD strategy, the test

This is the result that surprised me. Over the six years there were 88 daily closes below the lower band. On 78 of them the MACD line was already below zero. There were 89 closes above the upper band, and on 77 of them the MACD was already above zero. When price is at a band, the MACD agrees with it almost nine times out of ten.

It is not confirming anything — it is reporting the same twenty closes in a different shape.

The usual advice goes the other way: buy at the lower band only when the MACD is on the bullish side. I coded that too. Lower-band re-entry with the MACD line above zero, upper-band re-entry with it below — three trades in six years, all three lost, minus 553 pips. The setup almost never exists, because a close at the lower band is nearly always a close in a falling market.

Most articles present this filter as the safe version of the Bollinger Bands and MACD trading strategy. On my data it is the version that never trades.

What the histogram did do was thin the list. Band re-entry alone gave 74 trades and 1,490 pips. Asking the histogram to turn in my direction on the entry bar cut that to 51 trades and 1,046 pips. Fewer trades, less profit, about the same share of winners — 65% against 64%. A filter that removes trades without raising the win rate is not a filter. It is a delay.

Step by step trading plan

Readers ask for a step by step trading plan more than for anything else, so here is the rule exactly as my script ran it. I show it with pictures below, as a flow a robot can follow. The first step — settings. Bollinger Bands 20 and 2, MACD 12, 26 and 9, daily chart, nothing changed from the MetaTrader defaults.

The second step — wait for a daily close outside the lower band. Not a touch of the wick, a close. The third step — the next day must close back inside the band, and the MACD histogram on that day must be higher than the day before. If both hold, you buy at that close.

The fourth step — the stop. I put it one band-gap below the entry: the distance from the middle band to the lower band on the signal day. On 30 October that gap was 143 pips — $1,430 of risk on one standard lot, $14.30 on 0.01 lot, 14 cents on 0.01 lot of a cent account.

Lot size follows from the stop. If you risk 1% of a $10,000 account, that is $100, and $100 divided by 143 pips gives $0.70 per pip, so 0.07 lot. On a $100 cent account the same arithmetic gives 0.07 of a cent lot.

The fifth step — the exit. Close the trade when a daily close reaches the middle band, or after 15 bars, or at the stop, whichever comes first. The average trade in my test lasted seven days. Sells are the mirror: close above the upper band, next close back inside, histogram lower than the day before, stop one band-gap above.

Why a close and not a touch? Because a touch happens inside the day, and you cannot test it on closes. Bollinger’s own rule says the same thing in different words — the band is a relative definition of high and low, not an order to act. The close back inside is the market telling you the stretch is over. The histogram tick is a small extra, and after this test I treat it as optional.

What the backtest hides

I want to be careful here, because a table with plus signs sells itself. The data are noon buying rates from the Federal Reserve, one price per day. There are no highs and lows, so a stop that would have been hit inside the day counts as not hit. There is no spread. On 51 trades a one-pip spread costs 51 pips, which is small. There is no swap, and GBP/USD swaps over seven days are not zero.

The bigger problem is the shape of the result. Winners averaged plus 133 pips, losers averaged minus 187. The rule wins two trades in three and gives back more on the third. 2021 lost 390 pips, 2022 lost 139, and 2024 alone made 947 of the 1,046.

A rule that needs one good year in six is a rule I would not put real money behind without a second pair and a second timeframe. I ran it on daily closes of one pair. Whether it holds on H4, or on EUR/USD, I do not know — I have not run that yet.

There is a second way to read the same bands, and it is the one I lean to more — the double Bollinger Band strategy, where you watch the one-deviation bands and trade the zone between one and two deviations. It does not need the MACD at all.

If you are new to the bands, start with how Bollinger Bands work in forex. If the MACD part is the unclear one, the MACD explained page goes through the lines and the histogram one at a time.

Where the bands help a robot

None of the robots we sell were built around this rule, and after this test I am not going to add it. But the bands do one job well that the MACD cannot — they measure quiet. The October squeeze on GBP/USD was visible on 19 October, before the direction was.

A width reading under 1.6% on the daily chart has happened a few times a year on this pair, and each time the next three weeks moved more than the three before.

That is the filter I would put into a robot: trade the break of a squeeze, not the touch of a band. For the direction I use a trend tool that is not built from the same twenty closes — our Smart Trend Trading System for MT5 is free, and it is what sits on my own charts next to the bands.

So what did six years of GBP/USD say? Bands define stretch and quiet, the MACD repeats the bands in a different shape, the re-entry rule earns most of its money in one year, and the squeeze before the Bank of England rise was not only the clearest signal of the autumn but also the one no indicator confirmed.

The more of your own data you put under a rule before you trust it, the steadier your trading is going to be.

Data and sources: Federal Reserve, H.10 Foreign Exchange Rates, historical daily data; Bank of England, Bank Rate raised to 4%, 5 November 2026; John Bollinger, Bollinger Band rules; MetaTrader 5 help, Bollinger Bands and MACD; Gerald Appel, Technical Analysis: Power Tools for Active Investors, 2005.