XAUUSD Scalping Robot Settings for MT4 and MT5

You did not write the numbers your robot is trading on. Somebody you have never met did, on a broker you have never used, with an account size that is probably not yours. Those numbers are now running on your money.

That is the whole problem with XAUUSD scalping robot settings, and it is worth saying plainly, because most pages on this topic sell the robot instead of fixing the inputs. A gold scalper can be a reasonable tool. A gold scalper running someone else’s risk settings is a coin flip with a commission attached.

The problem: your XAUUSD robot is trading a stranger’s numbers

A settings file does not argue with you. It just lists values. RiskPercent=2, StopLoss=150, MaxSpread=0, LotMultiplier=1.7. Nothing in the panel explains that the third line switches off the filter that made the first line survivable, or that the fourth makes the second meaningless once a trade goes wrong.

You also cannot see, from a settings list, the three things that actually differ between your situation and the developer’s:

• Your broker’s spread. Gold is quoted with a spread that breathes. On a raw account in a calm London hour you might see 15 to 25 points. Ten minutes into a US inflation release, that same quote can be 100 points or wider, briefly, while the price is still moving. A robot targeting 400 points can absorb that. A robot targeting 150 points cannot.
• Your account size. A preset written for a $10,000 account produces a lot size that is four or five times too large on a $2,000 account. Nothing warns you. The robot simply risks more per trade than the author intended.
• Your latency. Scalping gold means competing for a price that exists for a second or two. A home connection and a broker server on another continent turn a planned entry into whatever price is left.

Add one more ingredient that is specific to gold, and the picture is complete. Gold is not a currency pair with a tight spread. One lot is 100 ounces, so a one-dollar move in the price is 100 dollars on a full lot. It gaps over weekends, it reacts to real yields, the dollar and risk appetite at the same time, and it can travel further in a thirty-second candle than in a whole quiet afternoon. On instruments like that, the settings do more work than the strategy.

Two traders can run the same expert advisor on the same day and get opposite results. The difference will not be the entry rule. It will be the stop distance, the risk percentage, the spread cap and the hours the robot was allowed to trade.

That is what this page fixes. Not which robot to buy, but the inputs that decide whether the robot you already have survives its first bad week, and how to check the arithmetic for yourself instead of trusting a file.

Agitate: nine days on a preset someone posted

The sequence below is an illustrative composite, not a real trader’s account. It shows a pattern that appears often.

Nadia, a composite trader, had 1,500 dollars in a live account she had funded slowly over a year. She picked up a free gold EA from a Telegram group, and the pinned message included a set file. The screenshots in that group were excellent. She loaded the preset on a cheap VPS and let it run.

Three properties of that file went unexamined, because nothing in the chat explained them:

1. MaxSpread=0 meant no spread filter at all in that particular EA. It entered on the signal, at whatever price the broker was quoting at that instant, including during data releases.
2. RiskPercent=3.0 with MaxPositions=4. Each signal could open up to four positions, and each was sized at 3 percent of equity from its own stop. Four positions is not diversification when they are the same instrument, in the same direction, opened inside ninety seconds.
3. The session filter was off. The robot traded the thin 01:00 to 04:00 window with exactly the same enthusiasm as the New York open.

Day one she was up 74 dollars and delighted. Day two flat. Days three to five bounced between plus 40 and minus 60, which is ordinary noise for a system holding several positions at once.

Day six was a US data morning. Spread on XAUUSD went from 22 points to a peak she never saw, because the robot took the trades anyway. Three positions opened in the same direction within roughly 90 seconds, each with a stop only two dollars wide. Gold turned and moved through all three stops inside four minutes. The slippage past each stop added a few more points, and because the positions were stacked the same price move closed all three at once.

Day seven, two trades in the thin Asian window cost 38 dollars in spread and commission before the market had decided anything. Day eight, another 55 dollars. On day nine she switched it off with 1,120 dollars left.

The bill: 380 dollars of trading loss, about 60 dollars of commission and swap, 11 dollars for the VPS, and nine days of checking her phone at traffic lights.

Nobody in that group was being dishonest. They shared a file that worked on their account. It did not transfer, because a preset is not a strategy. It is a set of decisions about risk and cost, and those decisions belong to the broker, the balance and the clock.

The expensive part came afterwards. Nadia’s next conclusion was that gold robots do not work. In fact, three inputs on that panel were wrong for her, and each one takes about two minutes to fix once you know what it should say. The rest of this page is those inputs.

What actually fixes a gold robot running the wrong settings

Five numbers decide most of the outcome: stop distance, risk per trade, maximum spread, session window and the position cap. Everything else in the panel is secondary, because those five determine whether you are still trading in three months.

The order matters, and it is the opposite of how most presets are built:

1. Stop distance first, because it is a property of the market. Gold’s noise level sets the minimum, not your preference.
2. Risk per trade second, because it is a property of your account. It should not change when you change brokers.
3. Position size third, because it is the output of the first two. This is the calculation the free bot on this page implements in its own position-sizing code, and it is the one most presets get wrong.
4. Maximum spread fourth, because it protects the edge you just sized. A spread cap is not a filter for bad traders; it is a filter on bad prices.
5. Session window and position cap last, because they decide how often the other four get tested. A robot with no session filter pays the worst spread of the day with the best risk settings in the world.

Start with the ranges below, then verify them against your own broker and your own demo log. A published preset, including anything on this page, is a starting point rather than an optimum. The optimum depends on data only you have: your spread history, your balance, the hours you can actually watch the screen.

How far should a XAUUSD scalping robot place its stop?

Wide enough that ordinary noise does not close the trade, tight enough that the loss is one you can repeat ten times in a row without changing your behaviour. On gold, that window is narrower than most presets assume.

Here is what different stop distances mean on a two-decimal XAUUSD feed, where 100 points equals one dollar of price:

Stop distance	In points	What usually happens
0.80 dollars	80	Noise. A quiet London hour can cover this in a single minute.
1.50 dollars	150	The common vendor preset. Works on M1 in calm conditions and fails on a normal data day.
2.00 to 3.00 dollars	200-300	The honest working range for M5 scalping on gold. Survives ordinary volatility.
4.00 to 5.00 dollars	400-500	Wide for scalping, usable on M15 or in a high-volatility regime.
8.00 dollars or more	800+	No longer scalping. Position size collapses and the target has to be huge.

Three rules sit on top of that table:

• Express the stop in price or in ATR, not in pips. On gold, “pips” means different things on different feeds. A dollar distance or an ATR multiple stays meaningful when you change brokers.
• Use ATR if the robot supports it. A stop of 2.5 times ATR(14) on M5 widens automatically when gold is excited and tightens when it is quiet. A fixed stop needs manual review every few weeks and usually does not get it.
• Check the broker’s minimum stop level. Some brokers refuse stops closer than a fixed number of points. A 60-point stop on a broker enforcing a 100-point minimum will be rejected, or quietly widened by the terminal.

If the panel offers a fixed 150-point stop and no ATR option, widen it to 250 or 300 points, re-run the backtest and compare the trade count. Most presets were written for a calmer gold market than the one you are trading.

What risk percentage per trade keeps a gold scalping robot alive?

Between 0.25 and 0.5 percent of equity per position for a first live run, and never more than 1 percent while you are still learning how the robot behaves. Accounts that die on gold are usually not killed by a bad entry rule. They are killed by a 2 percent risk setting meeting a five-loss run while the spread was widening.

The arithmetic that matters more than the percentage:

lots = (equity x risk percent) / (stop distance in price x 100)

Account equity	Risk per trade	Cash at risk	With a 2.50 dollar stop, the lot size is	What you can realistically trade
300 dollars	0.25%	0.75 dollars	0.003	Nothing. Only a 0.75-point stop fits the 0.01 minimum, and gold does not offer that.
1,000 dollars	0.5%	5.00 dollars	0.02	0.02 lots. Workable.
1,000 dollars	1.0%	10.00 dollars	0.04	0.04 lots, if the broker accepts micro lots in 0.01 steps.
5,000 dollars	0.5%	25.00 dollars	0.10	Comfortable. Trade 0.09 and keep the buffer.
25,000 dollars	0.5%	125.00 dollars	0.50	Spread cost becomes the main enemy, not lot size.

Read the first row again, because it is the most useful line here. At 300 dollars, a 2.50 dollar stop and the standard 0.01 minimum lot, you would be risking 2.50 dollars, which is 0.83 percent of the account before spread. That account has no room for a gold scalper at any risk setting. Either the balance grows, or the stop moves to a slower timeframe, or the instrument changes. No preset fixes it.

Two more rules belong in this section:

• Round the lot size down. If the calculation returns 0.037 lots, the answer is 0.03, not 0.04. Every platform rounds, and you get to choose the direction.
• Cap the total, not just the trade. Four positions at 0.5 percent is 2 percent of the account exposed at the same moment. Use the MaxPositions, MaxTotalRisk or MaxDailyRisk input, and set the daily figure low: a 2 percent day is a bad day, and 5 percent is a bad week that turns into a bad month.

What maximum spread should you set on XAUUSD?

A number you genuinely see during at least 80 percent of your intended trading hours. On most raw gold feeds that is 20 to 35 points. Expect the cap to block trades around data releases, and understand that blocking them is the point.

Spread is the rent your robot pays on every trade, and gold charges a lot of it. Read it as a share of the target rather than as an abstract number:

Spread at entry	Target	Spread as a share of the target	Verdict
15 points	400 points	3.8%	Fine. The edge survives.
25 points	400 points	6.3%	Normal for gold.
25 points	150 points	16.7%	Your broker is now the largest single counterparty.
60 points	150 points	40%	Stop trading. You are paying for the privilege of gambling.
100 points	400 points	25%	Only viable with a wide stop and a high win rate.

Ten minutes of work gives you your own number:

1. Open a XAUUSD M1 chart with a spread indicator, or watch the spread column in Market Watch.
2. Write the spread down at 09:00, 11:00, 14:30 and 21:00 in your broker’s server time, on a quiet day and on a data day.
3. Take the quiet-day value at your intended hour, add 30 percent, and use that as the cap.

If your broker’s average gold spread is above 40 points on a raw account, the honest conclusion is that scalping gold there is expensive. Either move the account to raw pricing or let the robot target a wider move on a slower timeframe. Our best MT4 EA round-up lists the execution conditions worth checking before automating on any broker.

Set a slippage value as well. MetaTrader’s default of 3 points is optimistic on gold; if the robot allows it, set slippage to 10 or 20 points and accept that some entries will be missed. A missed entry costs nothing, while a filled entry 15 points away from the planned price is a permanent loss of edge.

Which session windows should a XAUUSD scalping robot trade?

The London open through the London to New York overlap, and nothing else until you have a reason. Switch the robot off in the last hour of New York and through the Asian session unless the EA was written specifically for thin conditions.

Window (London time)	Gold liquidity	Typical spread	Suitable for scalping?
00:00-07:00	Thin	30-80 points	No. Ranges are small and the spread eats a large share of them.
07:00-09:00	Building	20-35 points	Marginal. Pre-London positioning creates false breakouts.
09:00-12:00	Deep	15-30 points	Yes. The best cost-to-range ratio of the day.
13:30-16:30	Deepest	15-35 points, spiking on data	Yes, with a spread cap in place. This is the overlap.
16:30-20:00	Thinning	25-50 points	Acceptable for trend entries, poor for scalping.
20:00-23:00	Thin	40-90 points	No. Small targets cannot survive that cost.

Two practical details when you enter these into the panel:

• Convert to server time. Most MetaTrader servers run on GMT+2 or GMT+3 with daylight saving. A robot set for “08:00 to 12:00” on a GMT+2 server in winter is trading 06:00 to 10:00 London. Do the conversion once, write it in a comment beside the settings, and redo it when the clocks change.
• Do not confuse a session filter with a news filter. They do different jobs. If the EA has a news filter, set the block window to 30 minutes either side of high-impact releases, and stay realistic about its limits: it handles scheduled events, not surprise headlines.

Volatility filters work the same way, and most presets are missing half of one. A minimum ATR value is standard. A maximum ATR value is what stops the robot taking 150-point stops into a market that is moving 200 points a minute. When gold’s M5 ATR triples, the settings that were sensible at 09:00 are no longer describing the same market. For context on which tools are worth loading on a gold chart at all, our guide to the best MT4 indicators for XAUUSD covers the short list that earns its screen space.

Which timeframe and entry settings belong in a gold scalping preset?

M5 is the sane default for a gold scalping robot. M1 works only with a spread cap under 25 points and a broker you can reach with low latency. Anything above M15 is a different strategy with a different risk model, and it deserves a different preset.

Timeframe	Signals per week (typical)	Average stop	What breaks it first
M1	100+	1.00-2.00 dollars	Spread, latency and any news spike.
M5	20-60	2.00-3.00 dollars	Loose spread caps and stacked positions.
M15	10-25	3.00-5.00 dollars	Session filters applied too narrowly; late entries.
H1	5-15	6.00-12.00 dollars	Not scalping. Different preset, different expectations.

If you change timeframe, change the stop with it. Copying an M1 preset onto an H1 chart is one of the most common ways a working robot becomes a losing one, because both the target and the stop were calibrated to the noise of a faster chart.

Do trailing stops and break-even help a gold scalper or hurt it?

They help in a trend and hurt in a range, which is another way of saying they cannot be optimised without knowing the future. Set them defensively and expect them to cost money on the choppy days that make up most of the calendar.

The reason is mechanical. A break-even trigger that moves the stop to the entry price as soon as the trade is 10 points up will be hit by gold’s ordinary retracement inside a move that is still valid. The result is a row of flat trades that would have been winners, and a win rate that falls without the average win rising to compensate.

A more honest configuration:

• Break-even: only after the trade is 1.5 times the initial stop in profit, not after a fixed 10 points.
• Trailing stop: a distance at least as wide as the initial stop, activated later. A 40-point trail behind a 300-point stop is a noisy exit dressed up as management.
• Partial close: if the EA supports it, bank a third at one times risk and let the rest run behind a trail. Gold’s winning moves are long when they happen, and this is the only management rule that reliably helps.

Test both variants in the Strategy Tester and compare the trade count, not just the final balance. If break-even removes 40 percent of the trades and improves the equity curve by 5 percent, you have learned something real about the market you are trading.

A copy-ready settings table for XAUUSD scalping

Use this as a starting point and then prove it on your own account. Every value is a range, because the correct number depends on your broker’s spread, your balance and the hours you can monitor.

Setting	Conservative start	Working range	What breaks when you push it
Risk per trade	0.25%	0.25%-0.5%	Above 1%: an ordinary five-loss run becomes a 10% hole.
Stop distance	2.50 dollars	2.00-3.00 dollars on M5	Under 1.50: noise stops you out. Over 5.00: position size collapses.
Take profit	4.00 dollars	3.00-6.00 dollars	Under twice the spread: costs eat the edge.
Maximum spread	25 points	20-35 points	At zero: no protection. Above 50: the robot barely trades.
Maximum positions	1	1-2	Above 3: correlated exposure, not diversification.
Maximum daily loss	2%	1.5%-3%	Above 5%: one bad day erases a month of work.
Session window	09:00-16:30 London	London open to New York overlap	Trading all day keeps the worst hours and the worst spreads.
News block	30 minutes either side	15-45 minutes	No filter: you trade the spike at the widest spread of the day.
Minimum ATR (M5, 14)	60 points	50-100 points	Set too high: the robot stops trading in usable conditions.
Maximum ATR (M5, 14)	400 points	300-500 points	No cap: calm-hour stops get hit within seconds.
Trailing stop	Off, or one stop behind	1-2 times the stop distance	Tighter than the stop: manufactured losses.
Break-even	1.5 times the stop in profit	1-2 times the stop	A fixed 10 points kills valid trades.
Magic number	Unique per chart	Unique	Duplicates: two robots manage the same position.
Slippage	15 points	10-25 points	Zero: MetaTrader either rejects the order or fills at any price.

Nothing in that table is exciting, and that is the point. These are the numbers that decide whether the account is still funded in six months. The settings that feel clever are the ones that show up in the drawdown.

How do you test settings without losing money while you learn?

Two stages before real money: a one-year backtest on real ticks, then at least two weeks on a demo account using the exact settings you intend to run live. Then go live at half the risk you plan to use long term, and treat the first month as paid education rather than income.

The MetaTrader 5 documentation is unusually direct about why the first stage matters. It states that during testing the spread is not modelled but taken from historical data, and it warns that the cruder tick-generation modes can manufacture a “Testing Grail” — a smooth equity curve that a live account will not reproduce. If a gold scalper looks superb in the rough modes, the documentation’s own advice is to re-run it in the most detailed mode available.

A testing routine that filters out most bad configurations:

1. Backtest a year on real ticks on the symbol and period you will actually run. Check the trade count before the profit: under 100 trades is an anecdote, not evidence.
2. Read the drawdown in cash. A 12 percent peak-to-trough drawdown on a curve that ends higher is tradeable. A 45 percent drawdown is not, however it ends.
3. Break the result apart. If the entire profit comes from two weeks inside the sample, the settings are fitted to an event that will not repeat on schedule.
4. Run two weeks on demo and compare trade count, average duration and worst day against the backtest. Wide divergence means the backtest described a different market from the one your broker provides.
5. Start live at half risk and journal the three numbers you will need later: worst losing streak in the test, worst day in cash, and the widest spread you saw during your live hour.

Those three numbers are your stopping rule. They only exist if you write them down before the first trade.

What exactly is in this download?

The download on this page is the MQL5 Liquidity Sweep Bot, an open-source expert advisor by carlosrod723, published on GitHub under the MIT licence. It is a MetaTrader 5 robot that trades smart-money setups — fractal liquidity sweeps, order blocks and Fibonacci zones — with the kind of risk and session controls this page has been describing. It is not a preset you can trust blindly, but it is a working example of the settings in the table, in code you can read before you run it.

That makes it a reference for the settings on this page rather than a rival to your own robot. Where your robot only shows you a fixed lot and a preset, this one exposes how risk per trade, daily drawdown and the spread filter are actually implemented — so you can check whether your panel says what you think it says.

Component	Detail
Tool	MQL5 Liquidity Sweep Bot
Platform	MetaTrader 5 only, from the MQL5 source in the repository
What it does	Trades fractal liquidity sweeps and order blocks on MT5, with the risk and session controls built in
Main inputs	Risk percent per trade, daily drawdown limit, maximum spread, session and kill-zone hours, stop and target logic
Extra features	Multi-timeframe Fibonacci zones, order-block detection, partial exits, trailing stop, optional LSTM signal filter
Licence	MIT, which permits commercial use, modification and redistribution with attribution
Author	carlosrod723
Cost	Free

What does the Liquidity Sweep Bot show that a preset file cannot?

It turns the settings debate into code you can check in a few minutes. Instead of trusting a preset’s claim about risk, you read exactly how this bot converts a risk percentage into a lot size, how it caps the daily loss, how it blocks trades when the spread is too wide and how it restricts trading to the London and New York kill zones.

That closes the loop on every other section of this page:

• If its position-sizing code returns 0.02 lots for a 0.5 percent risk and you were about to trade 0.10, you now know your preset was written for an account ten times your size.
• If its daily drawdown check would have stopped trading after three losses, you now know how it treats the day your preset kept buying into.
• If its spread filter sits at 25 points, you now know what a working maximum-spread cap looks like in practice.
• If its session check only trades the kill zones, you now know how a robot that ignores the thin hours is wired differently from one that does not.

It also shows you what to look for in robots that hide their risk. Plenty of free gold EAs accept only a fixed lot size, with no drawdown or spread logic you can see. Read how this open-source bot wires those controls, then check whether your own robot’s panel exposes anything comparable — and if it does not, treat the fixed lot as the warning it is.

What it does not do

• It does not ship presets for your broker. Its defaults were built and backtested on the author’s own symbol and account, so treat every number as a starting point.
• It does not promise an outcome. No tool can. It exposes its risk logic; the market decides what happens next, and losing capital is a normal possibility on any leveraged instrument.
• It does not protect you from a bad spread. Its spread filter only blocks trades you tell it to block; get the maximum-spread value wrong and it will still trade into a widening quote.
• It does not remove the need for a demo run. Run it on a demo account for at least two weeks with gold’s real spreads before you judge any setting by it.
• It does not cover every gold broker cleanly. Its session and spread logic assume the server time and symbol naming in the code; verify those against your own account before you trust its filter times.
• It is not a finished product to run live unmodified. It is a reference you read and test, not a robot whose results you should take on faith.

If you would rather start from a commercial robot that already carries tested preset ranges, our own gold pages document the specifications and the settings they ship with — Zenith Matrix EA and Onix Stratos XAUUSD EA. The trade-off is symmetrical: you get preset ranges and support, and you give up the ability to read the code and set the numbers yourself. If readable code matters more, EA31337 Libre and the geraked MT5 expert advisors are two more auditable codebases worth testing, and the free download library collects the tools we have licence-checked.

Your next step, and what to check before a single live dollar

Take the settings table above and put it next to the preset your robot is running now. For each row, ask one question: which of these two numbers would I defend in front of somebody who has seen my account statements? Change the ones you cannot defend, starting with the risk percentage and the maximum spread.

Then download the open-source bot, attach it to a XAUUSD demo chart and read how your chosen risk translates into lots on your balance. If the number looks too small to be interesting, that is not a flaw in the tool. It is the arithmetic of trading gold with a two-dollar stop, and it is better to meet it on a demo account than in a live drawdown.

Keep the robot on demo for two weeks with the new inputs. Watch one data release and see whether your spread cap did its job. Log the worst day in cash while it is still hypothetical. When the demo behaviour matches your backtest on trade count and average duration, move to live with half the risk you eventually intend to use, and accept that a losing month is a normal part of running any system on a leveraged instrument.

If you would rather study grid mechanics before you change a single input, safe grid settings for MT4 and the gold EA low-risk checklist both start from the same premise as this page: the settings decide more than the sales page. When you are ready to test the idea on a chart, start with the open-source bot, then compare what its risk logic tells you with what your robot was about to do.

Trading forex and CFDs on gold carries a high risk of loss and is not suitable for everyone. Leverage magnifies losses as readily as gains, gold can gap over weekends, and past performance — including every backtest and every preset discussed here — does not indicate future results. Test on a demo account first and only ever risk money you can afford to lose.