Best MT4 Settings for Gold Scalping: 37 Tested, Only 5 Survived Both Halves and Double the Spread

Search for the best MT4 settings for gold and you get lists: Supertrend 10/3, RSI 14 at 70/30, a 9/21 EMA for scalping. None of them comes with a number, and none of them tells you whether the setting made money in a month when gold fell as well as a month when it rose.

This page does both. It follows my gold session-by-session test, which found when to scalp gold, and my gold vs EURUSD comparison, which found that on gold the spread is cheap and the edge is the problem. Settings are where that edge has to come from.

How were the settings tested?

The data is 13,736 5-minute bars of COMEX gold futures (GC=F), June 16 to August 26, 2026. Every result is in dollars per ounce after a $0.30 cost per trade, which is an assumption rather than a measured spread — so every setting was run again at $0.60.

To catch settings that only suit one kind of market, the data was cut in two on July 22, 2026. In the first half gold fell from $4,350.80 to $4,117.00. In the second it rose to $4,674.90. A setting that works in both a falling and a rising month is worth more than a setting with a bigger total from one of them.

The four groups:

• Supertrend — ATR period 7, 10, 14 or 20 and multiplier 1.5, 2, 3 or 4, always in the market, reversing on every flip.
• EMA cross — 5/13, 9/21, 12/26, 20/50 and 50/200, always in, reversing on every cross.
• RSI fade — period 7 or 14, levels 20/80 or 30/70, held for 30 or 60 minutes.
• ATR stop and target — enter on the EMA 9/21 cross, stop at 1, 1.5, 2 or 3 times ATR 14, target at 1R or 2R, closed after 4 hours if neither is hit. If one bar touches both, the stop is counted first.

Which gold settings held up in both halves?

Five settings out of 37. Two more were positive in both halves but fell below zero at $0.60: the ATR stop at 1.5×ATR with a 1R target, at -$0.04, and RSI 7 at 20/80 held 30 minutes, at -$0.17.

The EMA 50/200 cross has the largest number per trade, and it is not a scalping setting — 80 trades in ten weeks, and most of its profit came from the rising half. Treat it as a direction filter, not an entry.

The top ATR setting is the one I would actually trade. It made money going up and going down, and it made money on both sides: its 148 long trades averaged +$2.09 and its 160 short trades +$1.00. In a test where gold ended $324 higher, a short side that still made money is the best evidence on this page.

What does the ATR stop change?

The entry in all eight ATR rows is the same EMA 9/21 cross. Only the exit differs:

Compare that with the same EMA 9/21 entry and a plain time exit: in my session test it lost $0.46 per trade held for 30 minutes and made $0.56 held for 60. The exit, not the entry, moved this rule from losing to the best scalping result on the page.

There is a clear shape. A 1×ATR stop is too tight for gold — a median 5-minute bar of $4.10 against a median ATR of $4.62 means ordinary noise hits it. A 3×ATR stop is too wide: both 3×ATR rows finished within a few cents of zero. The middle, 1.5 to 2×ATR, holds all three of the strongest rows, and a 2R target beat a 1R target at every stop size.

Notice this is the opposite of EURUSD. In my fixed vs ATR stop test every ATR multiple lost after costs on EURUSD M5, because 2×ATR there was only 4.4 pips, and a fixed 15-pip stop won; my stop loss distance test found the same peak. On gold, 2×ATR is about $9 — far above the $0.30 cost — and the ATR stop becomes the best exit in the grid. If you want the stop to trail instead of sitting still, my ATR trailing stop guide covers the indicator.

Is Supertrend 10/3 a good setting for gold?

Dollars per trade after the $0.30 cost. The default 10/3 ranked fifth of 16 and lost $0.54 per trade over 348 trades. Two combinations were positive — 20/4 at +$0.69 and 14/4 at +$0.42 — and both only because of the rising half:

All 16 Supertrend settings lost in the falling half. 12 of 16 made money in the rising half. No setting changes that; the month does. Supertrend’s long trades made money and its short trades lost, in a test where gold went up. That is not a setting working. That is a trend being followed.

If you already use Supertrend on gold, my Supertrend guide for MT4 covers installation and how the line behaves on the open candle, and my ADX vs Supertrend comparison tests a filter on top of it.

What about EMA pairs and RSI?

The same pattern as Supertrend: every fast pair lost in the falling half and made money in the rising half. The 9/21 pair lost $0.41 per trade in the first half when traded always-in, and made $1.60 per trade in that same half with a 2×ATR stop and 2R target. Same entry. Different exit.

RSI fades lean the other way from trend settings: the two positive 60-minute RSI rows made all of their money in the falling half and lost in the rising one. The only RSI row positive in both halves is RSI 7 at 20/80 held 30 minutes, at +$0.13 — and it turns negative at a $0.60 spread. The MT4 default, RSI 14 at 30/70, lost $0.29 per trade at 30 minutes. For RSI period choice on EURUSD, see my RSI 7 vs RSI 14 test.

How to set up the winning gold settings in MT4

Everything below uses indicators that ship with MetaTrader 4, so nothing needs to be downloaded. If you have not added an indicator before, setting up your first MT4 chart covers the basics.

1. Open an XAUUSD M5 chart. Your broker may call the symbol GOLD or XAUUSD.
2. Add two moving averages. Insert → Indicators → Trend → Moving Average. First: Period 9, MA method Exponential, Apply to Close. Second: Period 21, same method, a different colour.
3. Add ATR. Insert → Indicators → Oscillators → Average True Range, Period 14. It opens in its own window below the chart.
4. Enter on a closed-bar cross. When the 9 EMA closes above the 21 EMA, buy; below, sell. Do not act on a cross that has not closed.
5. Set the stop at 2×ATR. Read ATR on the signal bar. At the test’s median ATR of $4.62, that is a stop about $9.24 from entry.
6. Set the target at 2R — twice the stop distance, so about $18.48 from entry at the median ATR.
7. Close by hand after 4 hours if neither is hit. That was the test’s time limit.
8. Size by the stop, not the lot. On most brokers one standard gold lot is 100 ounces, so a $9.24 stop risks $924 per lot, and 0.01 lot risks about $9.24. My position sizing table does the arithmetic.

For timing, my session test found the London–New York overlap was the worst session for fast gold rules and the New York afternoon the best. This setting was not re-optimised by session, and I would not narrow it further on ten weeks of data.

Frequently asked questions

What are the best MT4 settings for gold scalping?

In this test, an EMA 9/21 cross on M5 with a stop at 2×ATR 14 and a target at twice the stop. It made +$1.52 per trade over 308 trades after a $0.30 cost, was positive in both a falling and a rising month, stayed at +$1.22 at a $0.60 spread, and made money on both long and short trades.

What is the best Supertrend setting for XAUUSD?

None of the 16 held up. Every combination lost in the half when gold fell, and 12 of 16 made money in the half when it rose. The default 10/3 lost $0.54 per trade over 348 trades. The best total, 20/4 at +$0.69, lost $0.27 per trade in the falling half.

What RSI settings work for gold on the 5-minute chart?

Not the default. RSI 14 with 30/70 levels lost $0.29 per trade held 30 minutes. RSI 7 at 20/80 held 30 minutes was the only RSI setting positive in both halves, at +$0.13 per trade, and it went negative at a $0.60 spread. RSI did not produce a gold scalping setting worth trading on its own.

How far should a stop loss be on gold M5?

Between 1.5 and 2 times ATR 14 in this test. A 1×ATR stop was too close — the median 5-minute bar was $4.10 against a median ATR of $4.62 — and a 3×ATR stop left the trade to the time limit. At the median ATR, 2×ATR was about $9.24.

Are these gold settings the same on MT5?

The indicators — Moving Average, Average True Range and RSI — exist in both platforms with the same inputs, so the settings carry over directly. Supertrend is not built into either and has to be added as a custom indicator.

Could these results be curve-fitted?

Any grid of 37 settings will produce some winners by chance, which is why each was split into two halves with opposite market direction and re-run at double the cost. Only 5 passed both checks. For more on how optimisation misleads, see my in-sample vs out-of-sample test and optimised vs default settings.