Gold Pantheon Sovereign MT4

This is the MetaTrader 4 edition. It runs the same strategy, the same settings and the same exits as the MetaTrader 5 edition of Gold Pantheon Sovereign. The performance figures on this page come from the MetaTrader 5 real-tick tests, which use the more faithful tick data. The MT4 build was backtested in the MT4 Strategy Tester and compared with those results before release. MT4 tester figures differ somewhat, because MT4 uses its own tick timing and fill model.

Running several EAs on one account is not a portfolio. It is several systems fighting over the same margin.

Gold Pantheon Sovereign runs several complete, independent strategy books inside a single instance, sized together off one balance rather than each pretending the others do not exist. That is the whole product: the books do not know about each other, but the sizing does.

It is built for the trader who would otherwise run several gold systems side by side and discover, in the worst week of the year, that all of them wanted margin at once. Here that week is priced in from the start, because the risk budget is shared rather than multiplied.

The number, and the number beside it

Across 2019-2026, on real tick data, the standard setting turned $5,000 into $6.5 MILLION — with a worst drawdown of 36.7% and not one losing year.

Turn the dial up and the same trades produce $210.7 MILLION from $5,000 at 56.3% drawdown. Turn it off entirely — flat 0.10 lot, no compounding, every year weighted the same — and it still returns $258,044 on $10,000 (the flat-lot run starts from $10,000 rather than $5,000 — 0.10 lots on gold from a smaller account measures margin pressure, not the strategy).

That last figure is the one to weigh. It is the same edge with the arithmetic taken out.

The dial goes further than that. Pushed to its measured limit the same engine returns $5.95 BILLION at 75.2% drawdown — a figure this page publishes because it is real and refuses to recommend because it is not survivable by anyone who would have to watch it. It is the last screenshot in the set and it carries a warning band. The settings past it were run too: the next setting up runs to the very last bar and still fails, because it puts the account 94% under its own peak on the way — that is not a risk level, it is a coin flip, and they are not offered.

And it is not a recent-years result. Run across 2010-2026 — the whole modern history of gold, every one of its 17 years positive, not one parameter changed between them — the standard setting returns $22.72 BILLION from $5,000 at 38.1% drawdown, and the flat-lot run returns $415,472 from $10,000.

What it trades

Each book reads structure on its own timeframes and places resting stop orders around confirmed levels; unfilled orders expire and cost nothing. Stops, targets and trailing are set before a position exists. A single dial scales every book at once, so the trader chooses one risk level and the internal proportions stay exactly as they were validated.

What it will never do

Read this list twice, because it is the part most gold EAs cannot write.

• No martingale. Position size never increases after a loss.
• No grid. No ladder of orders added into a move against you.
• No averaging down. A losing position is never enlarged.
• No hidden recovery mode that quietly triples exposure to repair a bad week.
• No stop-loss-free positions. Every order carries a real, broker-side stop from the moment it is filled.

Those five tricks are how a gold robot manufactures a smooth equity curve and a five-star launch, and they are why so many of them end in a single afternoon. This range refuses all of them, which is exactly why the drawdown figures on this page are honest enough to publish.

One input. You choose the risk.

The same engine, the same trades, the same entries and exits underneath every row. The only thing that changes is one number in the settings. That is the whole idea: you decide how hard it works, and you can see in advance exactly what that decision costs.

XAUUSD · real ticks · 2019-2026 · 1:500

XAUUSD · real ticks · 2010-2026 · 1:500

XAUUSD · real ticks · 2010-2019 · 1:500

Drawdown is the worst relative equity drawdown — the deepest hole the account was ever in as a percentage of its own peak, open positions included. It is a harsher figure than the "maximal drawdown" most pages quote, and it is the one that describes what you would have watched.

The fixed-lot rows show no growth rate on purpose. The lot size never changes, so a percentage return shrinks every year purely because the balance it is measured against has grown — it would make the longer and better record look like the weaker one. What those rows are here to show is the profit factor and the losing-year count, which is where a flat lot tells the truth that a compounded curve cannot.

"Losing years" counts calendar years. The recent and full-history windows end 1 May 2026, so their last year is a four-month stub; the isolated early window ends 1 January 2019 and contains nine complete calendar years.

Every setting listed above survived the entire test window. That is not automatic at the top of a ladder — push risk far enough and the account is liquidated part-way through, after which the tester reports a small, tidy, entirely meaningless balance attached to a curve that simply stops. Those settings were found, measured, and are not offered.

The drawdown column is one path. Here is the distribution behind it.

Every drawdown figure on this page — anywhere, from anyone — is what happened once, on the history that happened. History could have dealt the same trades in a different order, and the honest question is not "how deep did it go" but "how deep could it reasonably go".

So that was measured too. Each setting's trade ledger was resampled 10,000 times in blocks, which keeps winning and losing streaks intact instead of shuffling them into a smoothness that never existed:

Read that table before you choose a rung. The realised drawdown on every setting sits between the 35th and 83rd percentile of its own distribution — which means this record has not yet had a genuinely bad run, and you should not assume you have bought one that cannot. The right-hand column is an estimate, scaled from what the resampling measures onto the open-position equity your broker actually margins you on; it is deliberately the harsher reading of the two.

And there is a time cost, not only a depth cost. At the Standard setting the longest stretch below a previous high was 174 days in the test. Across the resampled paths the median is 223 days, and 12% of them spend more than a full year underwater. A system that is working exactly as designed can still spend a year looking like it is not. That is the thing that makes people switch a profitable system off, and it is why it is printed here rather than discovered later.

This distribution is included because one historical path does not define the full range of possible outcomes.

Six settings. That is the whole interface.

You are buying a finished system, not a parameter set to optimise. The internals are fixed, tested, and sealed — what you control is how hard it works and how it identifies itself on your account.

Every row of the table above ships as a ready-made preset file. Each rung in the ladder - including the flat-lot one, and including the ceiling under its warning - is a .set you load in one click, generated from the exact run whose drawdown is printed beside it. Nothing to copy off this page by hand, and nothing you have to guess at.

There is no optimisation to do, no timeframe to choose, and no set file to hunt for. Attach it to an XAUUSD chart, load the preset for the risk you are willing to carry, and leave it alone.

Requirements

• Symbol: XAUUSD (gold). Any chart timeframe — the system uses its own.
• Account: hedging, any broker. A raw-spread or ECN account is strongly preferred; the published tests use live raw-spread data.
• Leverage: 1:500 is what the tests were run at and what the risk model assumes. Lower leverage restricts the upper risk levels.
• Account size: $600 is the bare minimum, at the lowest risk level. $2,000 or more is recommended. One lot of gold is 100 ounces, and even the broker's smallest lot needs room to breathe on a small account. The published simulations use $5,000 at 1:500.
• A VPS is strongly recommended. Resting orders need a terminal that is actually online when the level is taken.

The honest part

The compounded figures on this page are a demonstration of edge, not a promise of an account balance. A backtest compounds without friction — it fills every order at whatever size the arithmetic asks for, however large that size has grown. A real account cannot. Past a certain equity those position sizes exceed the volume actually available at the level, and fills the tester grants for free start arriving with slippage, which compounds against the account exactly the way profit compounds for it.

So read the largest numbers here for what they genuinely establish: the edge underneath them is strong, and it held its shape across the entire period without one parameter being changed. A strategy that produces those figures on paper is a strategy capable of making serious money at a size a real account can actually trade — which is precisely what the fixed-lot row is on this page to show, because that row does not compound at all.

Beyond that, the usual and the true: this is a simulation on one broker's historical data. Gold's character changes. Spreads widen in events. A live account will not reproduce a simulation trade for trade, and past performance does not predict future results. Trade at a risk level you can actually sit through.