MACD Settings for MT4: The Default 12/26/9 Finished 6th of 9

Search for MACD settings and you get the same three numbers everywhere — 12, 26, 9 — usually with a note that Gerald Appel chose them in the late 1970s and a suggestion to try 5/35/5 for faster charts. Nobody tells you what either does on your pair. So I tested nine combinations on 16,926 real EURUSD 5-minute bars, June 3 to August 26, 2026, with one entry rule and one exit for all of them, and I also worked out what MetaTrader 4 is actually drawing when you add MACD to a chart — because it is not what the guides describe.

How do you add MACD in MT4?

Two minutes, and the same path in MT5. If you have not set the chart up yet, my first chart guide covers the six decisions that come before this one, and my installation walkthrough covers getting the platform on the machine.

1. Open the chart and timeframe you intend to trade. MACD’s settings are counted in bars, so 12/26/9 on M5 and 12/26/9 on H1 are different indicators.
2. Insert → Indicators → Oscillators → MACD.
3. In the Parameters tab you get three fields: Fast EMA, Slow EMA and MACD SMA. That third label is the platform quietly telling you something important, and I come back to it in the next section.
4. Set Apply to to Close. Leave it there — applying MACD to Median or Typical price changes every number in this article.
5. In Levels, add a level at 0 if there is not one. The zero line is the only fixed reference MACD has.
6. Drag the indicator window’s edge to make it taller than the default. MACD’s values on EURUSD M5 are tiny fractions of a pip and the standard window flattens them into a line.

Why does MT4’s MACD look different from every screenshot?

Because it is a different drawing of the same idea. The classic MACD has two lines — the MACD line and its signal — with a histogram of the gap between them. MT4’s built-in version draws the MACD line as the histogram and adds one line on top, which is the signal. That is why the crossover everyone describes is hard to see: on MT4 you are watching a line cross a bar chart rather than two lines crossing each other.

There is a second difference hiding in that MACD SMA field. MT4 smooths the signal line with a simple moving average, while the standard definition and most other platforms use an exponential one. The values are close but not identical, so a signal can arrive a bar earlier or later on MT4 than on the same settings elsewhere. My test used the classic exponential signal line, which is what MT5, TradingView and the textbook all use.

If you want the familiar two-line display on MT4 you need a custom indicator rather than the built-in one — I went through installing and reading that version in my 2-line MACD guide for MT4.

Which MACD settings tested best?

All nine combinations were run identically: MACD line crossing its signal line, entry at the close of the crossing bar, exit four hours later, no stop, no target, one trade at a time, and every run started from bar 120 so the slowest setting had the same warm-up as the fastest. Sorted by result after a 1-pip round-turn cost.

The default came sixth. That is the headline, and it is less exciting than it sounds, because the honest reading of this table is that nine settings all landed in a band a pip and a half wide, and every one of them was negative once the broker was paid. The spread between best and worst is 1.75 pips a trade; the spread you pay is 1.00. Choosing settings is a rounding error next to choosing whether to trade this rule at all, which is the conclusion my full MACD crossover backtest reached from every other direction.

Which of the three numbers matters most?

The signal length, and that surprised me — it is the parameter most people leave alone. Holding the averages at 12/26 and changing only the third number:

A monotonic improvement of half a pip and six points of win rate, from one field. The direction fits everything else on this site: the slower the signal, the better it did. The same held for the top of the main table — 3/10/16 has the longest signal line of any combination I tried relative to its averages, and it finished first.

The fast pairs were the disaster. 6/13/5 traded the most, 322 times, and lost the most, 1.09 pips each. This is the same relationship my moving average crossover test found across eight pairs and my EMA against SMA comparison found across four: more signals, more loss, every time.

Can you filter for the strong signals?

The obvious next idea is to keep only the crossings where the histogram is large, on the theory that a wide gap means real momentum. I tested it two ways on the default settings at a four-hour hold.

Look at that middle row for a second and then at the one below it. Taking the strongest half turns the rule positive. Taking the strongest quarter — being more selective by the same measure — turns it negative again, worse than no filter at all. A genuine effect gets stronger as you tighten it. This one does not, so the +0.91 is almost certainly an accident of which 245 trades happened to land in that bucket.

That pattern is the single most useful thing in this article, and it has nothing to do with MACD. It is what over-fitting looks like from the inside: a filter that produces a great number in one slice and falls apart one notch further along. I built the whole in-sample against out-of-sample test around this failure, and my optimised against default settings comparison found the optimiser winning three of four tests and still losing to the spread — exactly what happened here.

What settings would I actually put in the box?

• Leave it at 12/26/9 unless you have tested otherwise. Sixth of nine sounds bad, but the spread between all nine is smaller than your trading cost, and the default is the one everyone else is watching.
• If you change one number, change the signal. 12/26/18 beat 12/26/9 which beat 12/26/5, cleanly, in that order.
• Apply to Close, and add the zero level. Two settings that cost nothing and change what you are looking at.
• Do not run MACD next to a 12/26 EMA crossover. The zero line is that crossover — 677 of 677 identical bars, which I show in my MACD against moving average crossover comparison.
• Do not use MACD as your trigger on M5. Every setting here lost after spread. If you keep it on the chart, keep it as context.

Frequently asked questions

What are the best MACD settings for MT4?

On EURUSD 5-minute bars, 3/10/16 and 8/17/9 were the two best of nine combinations, at +0.66 and +0.54 pips a trade before costs. Neither cleared a 1-pip spread. The default 12/26/9 came sixth at -0.36. The practical answer is that settings are not where MACD’s problem is.

Are 12/26/9 the best MACD settings?

Not on this data — five of the eight alternatives I tested beat them. But all nine finished inside a band 1.75 pips wide and all nine were negative after costs, so the default is a perfectly defensible choice.

What is MACD SMA in the MT4 settings?

It is the signal line length, and the label is literal: MT4 smooths its signal with a simple moving average rather than the exponential one the standard definition uses. Values differ slightly from MT5 and TradingView on identical settings.

Why does MT4’s MACD only show one line?

The built-in indicator draws the MACD line as a histogram and the signal line as the single line on top. The familiar two-line display requires a custom indicator, which I cover in my 2-line MACD guide.

What MACD settings are best for a 5-minute chart?

Everything in this article was measured on 5-minute bars, so the whole table is the answer for M5. The longer-signal combinations did better and the fast ones like 6/13/5 did worst, losing 1.09 pips a trade across 322 trades.

Should I only take strong MACD signals?

I could not make that work. Filtering to the top half of crossings by histogram size gave +0.91 pips, but tightening to the top quarter gave -0.53 — worse than taking every signal. A filter that reverses when you tighten it is not measuring anything real.

Does MACD work better on higher timeframes?

I have not tested MACD above M5, so I cannot tell you. What I can say is that within M5 the slower parameter sets consistently beat the faster ones, and that the same relationship held for moving averages, Stochastic and Parabolic SAR on this dataset — see my Parabolic SAR test and my Stochastic settings guide.