XAU Patient Trend

XAU Patient Trend 2.0 is an Expert Advisor for gold (XAUUSD). Version 2.0 is the same build, with the same rules and default settings, that the developer has been running on their own real-money account since November 2026. It combines four independent modules on one chart:

• London auction scalp: a short sell in the ten minutes before the London gold auction (10:21 to 10:31 London time), and a five-minute one before the afternoon auction. It trades only when the same trade was profitable after costs over the last 40 trading days.
• Night-range breakout and R-Breaker: two intraday systems on M5 that always close the same day.
• Trend modules: M30 and H4 trend entries (Keltner, Ichimoku, Supertrend, H4 Donchian, H4 pullback) that follow the trend with trailing stops. They open buys only by default.

Every position has a stop loss from the moment it opens. There is no martingale, no grid and no averaging down. The default settings are aggressive: the auction scalp risks 7% of the balance per trade, times a multiplier of up to 2. Please read the sections on risk and weaknesses before you buy. The Classic set of the input "Strategy set" keeps the H1 trend rules of version 1.2.

How it works

London auction scalp (gold symbols only). The LBMA Gold Price auction starts at 10:30 and 15:00 London time. In the MetaQuotes 1-minute data the gold price has tended to fall in the minutes before the morning auction: from 10:21 to 10:31 London the average move was -0.87 USD per ounce in 2025 and -2.45 USD in 2026 (January to October), before costs. In 2014 to 2019 it was close to zero, which is less than the trading costs. The EA sells at 10:21:00 London time and buys back at 10:31:00 (afternoon: 14:56 to 15:01). Before each auction it calculates what the same trade would have made on each of the last 40 trading days after a cost of at least 0.22 USD per ounce (or the live spread plus your commission, whichever is larger). It trades only when the t value of these 40 results (mean / standard deviation x square root of 40) is at least 1.25, and the lot is multiplied by t / 2, between 0.5 and 2. Protective stop: 4 x ATR(14) on M1. No trades on weekends, UK bank holidays or when the spread is above 0.60 USD. London time is taken from the computer clock (GMT), so the trade time does not depend on the broker's time zone; the input "Broker server clock" is only used to find the past auction minutes in the price history. Choose US dates for brokers whose server time follows New York (GMT+2 in winter, GMT+3 in summer, the most common setup), EU/UK dates for servers like the MetaQuotes demo. Over 2014 to 2019 the filter allowed only 86 auction trades, so the EA sat out most of the weak years.

Night-range breakout (M5, server time). The EA takes the high and low of 02:00 to 07:00 server time. The range is used only if its width is between 0.10 and 0.80 x ATR(14) on D1. Only the side that the price breaks first counts (5% of the width outside the range), and only in the direction of the D1 trend (EMA 50 above EMA 200 and the close above EMA 50 for buys, the opposite for sells). Stop: the other side of the range, at most 0.8% of the price away. No take profit, no new entry after 21:00, always closed at 22:00 server time, no trades on Wednesdays.

R-Breaker (M5, server time). Six levels from the previous day's high, low and close: breakouts beyond the outer levels and reversals after a failed test, both directions, decided on closed M5 bars. Emergency stop 0.5 x ATR(14) on D1, no take profit, always closed at 22:45 server time.

Trend modules (buys only by default). M30 Keltner channel breakout (EMA 20 +/- 2.5 x ATR), M30 Ichimoku Tenkan/Kijun cross, M30 Supertrend (10, 4) flip, H4 Donchian 30 breakout and an H4 pullback entry (a reversal bar in the 38.2% to 61.8% retracement of the last swing). They trade only in the direction of the H4 trend (and the D1 trend for all but Ichimoku). Initial stops from 1.5 to 4 x ATR (at most 10 USD on M30 and 25 USD on the H4 Donchian; the pullback entry uses the pullback extreme), then ATR trailing stops, no take profit. The M30 modules add one more position when a trade is +5R at a bar close (1.5 x the normal risk) and do not open on Wednesdays. The Keltner module uses half the normal risk.

Position size. The trend and intraday modules risk 1% of the balance per trade; the auction scalp risks 7% x the multiplier. The lot comes from the broker's contract values and is rounded down to the lot step. The EA never skips a trade because of the minimum lot, so on a small balance the minimum lot can risk much more than these percentages. If the free margin is not enough (low leverage), the auction lot is reduced to fit.

Drawdown guard. When the balance falls 15% below its highest value, every module risks half until the balance is back within 7% of the peak. Deposits and withdrawals move the peak by the same ratio, so a withdrawal is not counted as a loss.

Recommended setup

• XAUUSD, H1 chart (the modules read M1, M5, M30, H4 and D1 data themselves). One chart per account.
• A low-cost account (raw spread plus commission). The auction scalp needs total costs of roughly 0.30 USD per ounce or less; with higher costs it trades less or not at all. Enter your round-turn commission per ounce in the input (6 USD per lot = 0.06).
• A hedging account, because several modules can hold positions at the same time. On a netting account the auction scalp does not trade while another position is open.
• Balance: 1,000 USD or more. Leverage 1:100 or higher.
• A VPS or a computer that stays on, with a correct clock. Broker server time GMT+2/+3 is recommended for the intraday modules (they were built and tested on such servers).

Backtest summary (MT5 Strategy Tester backtest, not live trading)

Test conditions: XAUUSD H1, initial deposit 1,500 USD, leverage 1:100, swaps included, no commission, MetaQuotes-Demo data (variable spread), default inputs with "Broker server clock" = EU/UK dates (the dates of the MetaQuotes server). Long tests in 1-minute OHLC mode (99% history quality), the last 12 and 24 months with real ticks. With compounding on, the tester grows the lot without limit and the balance reaches millions within two years; such lot sizes are not realistic at a real broker. The main results below therefore use compounding off (lots sized on the 1,500 USD start).

January 2014 to November 2026, compounding off: 1,500 to 45,351 USD, profit factor 2.15, maximum balance drawdown 17.3%, maximum equity drawdown 18.0%, 5,078 trades, 46.1% winning. A commission of 6 USD per lot round turn would have cost about 2,900 USD. Without the auction scalp the same test ended at 14,829 USD (profit factor 1.69, maximum equity drawdown 24.7%).

Profit per calendar year (compounding off, USD): 2014 +271, 2015 -233, 2016 +564, 2017 +168, 2018 -128, 2019 +311, 2020 +9,891, 2021 +1,982, 2022 +905, 2023 +673, 2024 +3,095, 2025 +9,244, 2026 (January to November) +17,109.

Last 12 months (real ticks): compounding off 1,500 to 19,661 USD, profit factor 2.98, maximum equity drawdown 23.6%, 651 trades. With compounding: 2,506,006 USD, maximum equity drawdown 26.7%, largest single loss 12.9% of the balance.

Last 24 months (real ticks, history quality 73%): compounding off 1,500 to 27,194 USD, profit factor 2.77, maximum equity drawdown 13.4%. With compounding: 7,461,913 USD, maximum equity drawdown 26.5%.

2014 to 2019 with compounding: 1,500 to 2,460 USD in six years, profit factor 1.12, maximum equity drawdown 21.1%.

By module (2014 to 2026, compounding off):
- London auction scalp: 1,189 trades (0 buys, 1,189 sells), 64% winning, profit factor 2.69, net +30,657 USD.
- Night-range breakout: 801 trades (558 buys, 243 sells), 45% winning, profit factor 1.34, net +1,671 USD.
- R-Breaker (M5): 1,706 trades (918 buys, 788 sells), 44% winning, profit factor 1.20, net +1,420 USD.
- Trend modules (M30/H4): 1,382 trades (1,382 buys, 0 sells), 34% winning, profit factor 2.26, net +10,103 USD.

Classic set (H1 trend only, 1% risk), January 2020 to November 2026, compounding: 1,500 to 4,808 USD, profit factor 1.57, maximum equity drawdown 16.3%.

How the settings were chosen. The auction effect was found in October 2026 in 2014 to 2026 data, and its 40-day filter, cost rule and multiplier were chosen while looking at 2020 to 2026. The trend modules, the module combination, the buy-only rule and the risk levels were chosen in October and November 2026 while looking at the same backtests shown here. These results are therefore largely in-sample and are likely to look better than future results.

These are backtest results. They are not live trading results and they are not a forecast. Prices, spreads, commissions, swaps and execution at your broker will lead to different results. You can test the EA yourself with the free demo in the Strategy Tester.

Risks and weaknesses

Aggressive default risk. One auction trade can lose about 7% to 14% of the balance. In the last 12 months with compounding the largest single loss was 12.9%. Lower "Auction scalp: base risk" (for example to 2 to 4) for smaller swings.

The auction edge may weaken or disappear. It was strong in 2020 and especially in 2025 and 2026; in 2022 to 2024 the average move was smaller than typical costs. It is a timing pattern with no proven cause. If the auction organiser changes the auction time or method, or if many traders act on it, it can fade. The 40-day filter stops trading when it fades, but only after some losing days.

Costs. The auction trades target a move of about one to two USD per ounce. A wide spread, a high commission or slow execution removes a large part of it.

Weak years. 2014 to 2019 made little, with losing years 2015, 2018 (compounding off). Most of the profit came in 2020 and in 2025 to 2026, when gold moved the most.

Trend modules buy only by default. In a long falling market they do not sell; the intraday modules and the auction scalp still trade both ways or short.

Several positions at once. A bad day can hit more than one stop. Gaps after weekends or news can fill stops at worse prices.

Small balances. The minimum lot is always traded, so the real risk on a small balance can be far above the risk setting.

Server time. The intraday modules use broker server time; on a server that is not GMT+2/+3 their windows move (not tested).

Short live record. The developer's live account has run this build only since November 2026.

Input parameters (names as on the Inputs tab)

Strategy

• Strategy set: 2.0 (default) or Classic (the H1 trend module of version 1.2, with a 2 x ATR initial stop).

London gold auction scalp (gold only)

• Auction scalp: sell 10:21 -> buy back 10:31 London (AM) — default true.
• Auction scalp: also the PM auction (14:56 -> 15:01 London) — default true.
• Auction scalp: base risk per trade (% of balance), x the multiplier — default 7.0.
• Auction scalp: max lot multiplier (t / 2, at least 0.5) — default 2.0.
• Auction scalp: trade only when the last 40 days t value is at least this — default 1.25.
• Auction scalp: your round-turn commission per ounce (6 USD per lot = 0.06) — default 0.06.
• Broker server clock: daylight saving dates (for past auction times) — default US dates.

Intraday modules (server time, flat the same day)

• Night-range breakout (02-07 range, trend direction, flat 22:00) — default true.
• R-Breaker (M5, flat 22:45) — default true.

Trend modules

• M30 Keltner/Ichimoku/Supertrend + H4 Donchian + H4 pullback — default true.
• Trend modules: buy entries only — default true.

Money management

• Risk per trade, trend and intraday modules (% of balance), 0 = fixed lot — default 1.0.
• Compounding: current balance (true) / start balance (false) — default true.
• Fixed lot (used when risk % is 0) — default 0.01.
• Max spread in price units (0.60 = 60 cents on gold), 0 = off — default 0.60.
• Max slippage in price units (0.50 = 50 cents on gold) — default 0.50.
• Magic number (each module adds its own offset) — default 20260927.

Drawdown guard

• Cut the risk after a drawdown from the peak balance — default true.
• Guard starts at this % below the peak balance — default 15.0.
• Risk multiplier while the guard is on — default 0.5.
• Back to normal risk within this % of the peak — default 7.0.

Installation

1. After buying or renting, the EA appears in the MT5 Navigator under Expert Advisors, Market.
2. Open an XAUUSD chart and set the timeframe to H1.
3. Drag the EA onto the chart. On the Common tab, allow Algo Trading. On the Inputs tab, check the auction risk, your commission per ounce and the broker server clock.
4. Turn on the Algo Trading button. When the EA starts, it writes its settings to the Experts log. Before each auction it writes its decision (t value and lot multiplier) to the log.
5. Keep the terminal running. After a restart the EA continues with its stored peak balance.

Frequently asked questions

I used version 1.x. What changes?
The default strategy is now the 2.0 set. Choose "Classic" in "Strategy set" to keep the version 1.2 trend rules (the default risk is now 1%; it was 2% in 1.x). Open positions of 1.x with the same magic number keep being managed.

Which broker?
Any MT5 broker with XAUUSD, low costs and a GMT+2/+3 server. The auction scalp switches itself off on symbols that are not gold.

Will my results be the same as the backtest?
No. Results depend on your broker's prices, costs, server time and execution, and on future markets.

Can I run other EAs on the same account?
Yes. The EA manages only positions with its own magic numbers (the input number plus a fixed offset for each module).

Can I run it on a prop firm account?
The EA has no daily loss limit and the default auction risk is high. Check the rules of your account first.

Support is provided through the product comments and MQL5 private messages.

Trading gold with leverage carries a high risk of loss. Past performance, including backtests, does not guarantee future results. Use only money you can afford to lose.