EXPERT ADVISOR · METATRADER 5 · GOLD ONLY

An XAUUSD expert advisor that only knows gold

Most gold EAs are an FX system pointed at a metal. This one was built, tested and cut down on XAUUSD data alone. We tried 34 strategies, shipped 4, and deleted a whole second instrument after it lost money over 117 trades.

What separates a gold EA from a generic one

Stops sized to gold’s noise floor

Gold can move several dollars without meaning anything. A stop calibrated for an FX pair sits inside that noise and gets taken out by movement that was never a reversal.

A spread ceiling set in gold dollars

Our threshold is one dollar on XAUUSD. Copy that number to a different instrument and it blocks either every trade or none, which is how a spread filter ends up doing nothing.

Session structure it can use

Gold trades in sessions, and each one leaves an opening range, a high, a low and a known pattern of liquidity. Several strategies key off those levels, and on a 24/7 instrument they have nothing to hold onto. When we replayed the bot from Aug 2025 to early Oct 2026 trading 24 hours instead of 06:00 to 21:00 UTC, its 60 extra night trades lost money as a group.

XAUUSD EA questions

What is an XAUUSD EA?

An XAUUSD EA is an Expert Advisor built to trade gold against the US dollar. An Expert Advisor is an automated program that runs inside MetaTrader: it watches price, applies coded entry rules, and places and manages trades while you are away from the screen. The XAUUSD part matters because gold moves differently from an FX pair, and an EA tuned for EURUSD behaviour usually does badly on it.

Why does gold need a dedicated expert advisor?

Every parameter in an EA encodes an assumption about the instrument. Stop distances assume a noise floor, spread filters assume a typical spread, and session logic assumes when the market is liquid. Gold routinely moves several times the daily range of a major FX pair, so a stop that sits safely outside routine movement on EURUSD sits right inside it on XAUUSD. We learned this the expensive way when we ran the same engine on Bitcoin: 117 trades at a 28% win rate, and then we removed the symbol.

What timeframes does it trade?

It scans M1, M5 and M15 and reads market structure from a higher timeframe for directional bias. The lower timeframes give gold enough movement to trade intraday, and the higher-timeframe read stops the bot trading against the dominant direction.

Does it work with any MT5 broker?

Any MetaTrader 5 broker offering XAUUSD. It is built and tested on Exness, and nearly every client runs an Exness Standard account. On Exness, gold is XAUUSDm on a Standard account and XAUUSD on Pro, Raw Spread and Zero accounts, and since version 1.0.204 the bot trades whichever name the account has. At other brokers, confirm the gold symbol name matches and check the typical spread. The strategy mix is spread-sensitive, and the four strategies skip an entry while the spread is above their limit, $1.00 by default.

How many trades does it take?

The bot itself takes very few. Its gate chain checks regime, session, spread, cooldown and concurrency, and it rejects far more setups than it accepts. Some days it does not trade at all. In the 22-month backtest the whole profile took 1,297 trades, about 59 a month, but only 112 of them came from the bot; the three clock rules on gold took the rest. If you want an EA that is always in the market, this is the wrong one.

Is it a martingale or grid EA?

No. In the default and Conservative profiles every position is a fixed 0.02 lot (0.01 for the overnight rule under a $200 balance), the same after a loss as after a win. Under Custom the size comes from the balance unless a fixed lot is set. It never averages down, never adds to a losing position and never opens recovery trades. Prop firms ban those strategy types more often than any others, which is why this one is usable on a funded account and many gold EAs are not.

What results has it produced in testing?

On live client accounts, the bot's accuracy has been 60–80%. In backtesting, we replayed the Conservative + gold rules profile at a fixed 0.02 lot on 22 months (Jan 2025 – Oct 2026), each month on its own from $1,000 with a fresh bot. The average month made +$274.60 and 17 of 22 months were profitable. The worst month lost $575.12, and the worst drop inside a month was $860, 86% of the account. Across 1,297 trades the win rate was 47%. Do not plan on the average: two months, October 2025 and January 2026, made 56% of the total, and the other 20 averaged +$132.86. Earlier years were much weaker: on 2019–2024 price history the same profile made about +$5 a month, close to break-even. The backtest figures are hypothetical results, not a live track record.

The interface

Five of the ten screens. The complete walkthrough covers the backtester, shadow trades and pattern intelligence too.

Tech Kick Bot desktop dashboard showing the live log, open trades and the strategy performance tracker

Dashboard

Start and stop the bot, see open positions and read the live log as it works. Every scan prints the regime it detected, the bias, the spread it measured and whether it acted. When the bot sits still, the log names the gate that stopped it, so you can tell "it is broken" from "there was no setup".

Bot configuration screen with MT5 credentials, risk per trade, reward-to-risk ratio and market regime selection

Settings and configuration

Where you connect the MT5 account and set the risk. Risk per trade and reward-to-risk decide how long an account survives, so both sit on sliders on this screen. The regime checkboxes at the bottom choose which market conditions the bot may trade at all. Chop ships switched off because it did not earn money in testing.

Daily trading protection screen with configurable maximum daily profit and loss limits

Daily protection limits

A daily profit limit and a daily loss limit. Once either is reached the four strategies stop trading until the broker's server midnight. You cannot argue with it in the moment, and the moment right after a loss is when the urge to trade it back is strongest. On a prop firm account this is the control for the daily drawdown rule: set it below the firm's limit, never equal to it, and use Custom settings or the Conservative profile. A new install starts on Conservative + gold rules, and its three gold clock rules are not held by the daily loss limit, so change the profile first. Custom and Conservative run without those rules, so the limit applies to every trade. Conservative fixes the lot at 0.02 and the limit at $40. Under Custom the bot sizes each trade from the balance unless a fixed lot is set, so set the lot and the limit yourself before the first trade.

Market regime screen showing seven market state dimensions and per-strategy fit scores

Market regime detection

Reads current conditions on seven dimensions: direction, speed, volatility, momentum, liquidity, event risk and higher-timeframe alignment. The seven resolve into one bucket. It exists because trend systems bleed in ranges and mean-reversion bleeds in trends. The fit score beneath decides which strategies may compete for the next trade at all.

Strategy Lab table showing per-regime win rate, Wilson lower bound and flip state for each strategy

Strategy Lab

Each strategy's record in each regime, with a switch to disable or flip any cell. Win rate misleads on small samples, so the WLB column shows the Wilson lower bound, the pessimistic end of the confidence interval. The engine ranks on that number, so a strategy that won its only two trades cannot outrank one with a long record.

Account numbers, balances and profit figures are blurred in every screenshot. An account number plus a server name is most of an MT5 login, and a screenshot of one account over a handful of trades is not evidence of anything. The backtest figures and the full strategy record are published separately, with their sample sizes attached.