Petko AleksandrovDisclosure: some links here are affiliate links, and we earn a commission if you buy through them at no extra cost to you. Every figure comes from accounts we fund and monitor ourselves, never from a customer’s account and never from vendor marketing.
Both work gold, so they compete for the same buyer. They are not the same kind of machine. Prime Scalper opens one position at a time, holds it for minutes, and closes it when wrong, booking real losses along the way for a 19% realized drawdown while its open positions never floated more than 3% underwater across 8 months. Dark Gold runs a martingale that increases size every time it is losing, and across 9 months it booked almost nothing, a 4% realized drawdown, while its open positions floated to 9%. Dark Gold wins more often. Prime Scalper’s tracked accounts finished further ahead.
| Prime Scalper | Dark Gold | |
| Mechanism | Scalper, one position | Martingale, adds to losers |
| Max floating loss | 3% | 9% |
| Realized drawdown | 19% | 4% |
| Profit factor | 7.15 | 1.80 |
| Return | +43% | +7% |
| Accounts in profit | 2 of 2 | 3 of 3 |
| Record length | 8 months | 9 months |
If you run a funded account with a hard loss limit, the mechanism row decides it before you read another word, and the reason is not the floating-loss figure you see today.
What an expert advisor actually does
Worth a short detour, because the mechanical difference between these two products is the entire comparison.
An expert advisor is a program running inside MetaTrader that places trades according to fixed rules, without you clicking anything. It reads incoming quotes, evaluates whatever conditions its author wrote, and acts. Most read some combination of levels and indicator values: a moving average crossing, a volatility measure widening, an oscillator reaching an extreme. The specific indicator set rarely matters as much as buyers assume.
What matters far more is what the program does when it is wrong. That single design choice separates the entire category into groups, and it is where almost all the danger lives.
Three broad approaches cover most of what you will meet:
One position, closed when wrong. The program takes its loss and waits for the next setup. Losses are visible, frequent and small.
Several positions at the same size. As the market moves against the first entry, more open alongside it, all equal, and the group closes together when the market comes back.
Several positions at increasing size. Same idea, except each new entry is larger than the last, so a smaller reversal returns the group to profit.
Prime Scalper belongs to the first group. Dark Gold belongs to the third, which is the one traders most often misjudge. Everything else in this comparison follows from that.
Across the 58 products we can report on, 49 use one of the two multi-position designs. Seven hold a single position at a time, with two more running portfolios of single-trade strategies. If that ratio surprises you, it surprised me too when we first counted.
What backs these figures
We monitor 516 funded accounts holding 277,277 closed trades. Records reach 31 months at the longest.
Everything below is measured from executed trades on accounts we fund. Nothing simulated, nothing supplied by a seller, and the losing accounts are published beside the winning ones. The underlying data sits on our tracker.
Two measurements carry the comparison, and confusing them is the most common mistake I see:
Realized drawdown is the deepest fall in a robot’s own run of closed trades, measured against the account’s high-water mark. It counts only that robot’s booked results.
Max floating loss is how far underwater the unclosed positions went at their lowest point. The second one is what exposes a multi-position design, because a program can show a serene account curve while quietly holding an unrealized loss it has not yet booked, and the gap between a robot’s realized and floating figures is where its true temperament shows.
Prime Scalper: what our tracking shows
| Measure | Result |
| Instrument | Mostly gold |
| Mechanism | Scalper, one position at a time |
| Typical hold | Minutes |
| Entries per day | 1.24 |
| Win rate | 69% |
| Adds to losing positions | No |
| Realized drawdown | 19% |
| Max floating loss | 3% |
| Profit factor | 7.15 |
| Return | +43% |
| Accounts in profit | 2 of 2 |
| Tracked for | 8 months |
Loading trading accounts…
Three percent is the tightest floating-loss figure anywhere in our metals group. For scalping a metal this volatile, that is unusual enough that I checked it twice, and the 7.15 profit factor beside it says the winners clear the losers by a wide margin rather than scraping past them. Note the honest counterweight in the same table: the 19% realized drawdown. This product takes real losses and books them, which is why you can see its risk on the balance line rather than hidden in open positions.
The caveat is sample size. Two accounts over eight months is thin, and a strategy that has not yet met a hostile stretch always looks tidier than one that has. Read this as encouraging rather than settled.
Dark Gold: what our tracking shows
| Measure | Result |
| Instrument | Mostly gold |
| Mechanism | Martingale, increases size on losers |
| Typical hold | Hours |
| Positions open at once | Several |
| Win rate | 85% |
| Adds to losing positions | Yes, every time |
| Realized drawdown | 4% |
| Max floating loss | 9% |
| Profit factor | 1.80 |
| Return | +7% |
| Accounts in profit | 3 of 3 |
| Tracked for | 9 months |
Higher win rate, all three tracked accounts in profit, and a calm-looking set of risk figures: 4% realized, 9% floating. That calm is the point of the next section, because it is produced by the mechanism rather than by the absence of risk. A martingale books very little precisely because it declines to close its losers, parking the strain in open positions and, when a trend refuses to turn, in the tail that has not arrived yet.
Why Dark Gold’s calm numbers are the ones to distrust
Here is the part that only becomes visible when you measure behavior rather than read descriptions.
Read the two products’ risk columns side by side and Dark Gold looks like the safer machine: a 4% realized drawdown against Prime Scalper’s 19%. That reading is exactly backwards, and the mechanism explains why. Prime Scalper’s 19% is booked, real, and finished; it took those losses and moved on. Dark Gold’s 4% is small only because it refuses to close a losing basket, so the loss sits open at 9% rather than being realized, and its floor is set by the market turning rather than by any rule inside the program.
The number that matters is not what Dark Gold has floated to in nine months, but what its category does when the run goes wrong. Across our full dataset, the worst account decline recorded by a martingale reaches 91%. Dark Gold’s nine months simply have not met that yet, and a low realized figure from a design like this is deferred risk, not absent risk.
Which is why we classify products by what their executed trades do rather than by the adjective in the title. Dark Gold’s orders increase size on losing positions, every time, so we class it a martingale regardless of how it is marketed. Before buying anything in this category, open the trade history yourself, look at whether position sizes increase after a loss, and trust that over any label. Our guide on spotting EA problems covers the rest of the checks.
Head to head: how they operate
| Prime Scalper | Dark Gold | |
| Strategy type | Scalper | Martingale |
| Markets traded | Mostly gold | Mostly gold |
| Typical holding time | Minutes | Hours |
| Trades per day | 1.24 | Not separately tracked |
| Positions open at once | One | Several |
| Adds to losing positions | No | Yes, every time |
| Best suited to | Hard loss limits, tight exposure requirements | Accounts that can absorb a martingale’s tail |
Head to head: what they did
| Prime Scalper | Dark Gold | |
| Tracked for | 8 months | 9 months |
| Accounts tracked | 2 | 3 |
| Win rate | 69% | 85% |
| Return | +43% | +7% |
| Realized drawdown | 19% | 4% |
| Max floating loss | 3% | 9% |
| Profit factor | 7.15 | 1.80 |
| Accounts in profit | 2 of 2 | 3 of 3 |
Three percent against nine, and why the gap understates the risk
Percentages stay abstract until you convert them into an account you would recognize.
Take a $10,000 account. Running Prime Scalper at its worst recorded moment, the open position was showing roughly $300 underwater. You would notice it and then forget about it.
Same account running Dark Gold at its worst so far: about $900 underwater, held across hours while the group waited for the metal to come back. That, on its own, is not alarming, and this is precisely the trap. The nine-month figure is not the ceiling. A martingale adds size to a losing basket, so the open loss grows as the move extends, and the category’s worst account in our records fell 91% before it was done. The honest way to read Dark Gold’s 9% is as the deepest it has been tested to, not the deepest it can go.
Three consequences follow, and they matter more than either return figure:
Margin becomes the binding constraint once the tail arrives. With several positions open and size increasing on each losing addition, whether you survive depends on your capital and your account terms, not on whether the strategy is eventually right.
Adding anything else becomes dangerous. A second program opening positions while a martingale is deep in a recovery can push combined exposure past the point of return.
The urge to intervene peaks at exactly the wrong moment. Closing at the bottom turns a temporary loss into a permanent one, and I would guess that behavior has cost our readers more than any purchase they ever made.
The last point is why the deciding factor here is rarely the returns column. Dark Gold’s +7% belongs to accounts whose recovery arrived inside the window we tracked. The mechanism offers no guarantee that the next one will.
Why the higher win rate belongs to the riskier product
Dark Gold wins 85% of its trades. Prime Scalper wins 69%. Read casually, that makes Dark Gold look like the safer purchase.
It works the other way, and the mechanism explains why.
A program that increases size on losers wins more often almost by construction. Each additional position lowers the average entry, so a smaller reversal returns the whole group to profit. The wins get booked and counted; the loss sits open, uncounted, until either the market comes back or the account cannot hold it. High win percentages in this category are usually a symptom of that design rather than evidence of a better edge.
Our records contain a blunter example. One product elsewhere in the dataset wins 98% of its trades and is a martingale underneath, booking that stream of wins while a losing basket sits open behind them. That is not an exotic edge case; it is what the design does, and the rare loss is where the account goes.
So the rule I would suggest: never read a win rate without an exposure figure beside it. Alone it tells you almost nothing, and in this category it frequently tells you something misleading. Our martingale explainer works through the arithmetic properly.
What each one feels like when it is going badly
Tracked records tell you what happened. They rarely tell you what it was like to sit through, and that gap is where most buyers make their actual mistake.
A bad stretch with Prime Scalper is visible and dull. Positions close, some at a loss, and the account curve drifts down in small steps toward that 19% realized figure. Nothing hides. After a few weeks of that you start asking whether the edge has stopped working, because there is no dramatic moment to point at, just erosion. Plenty of people abandon a functioning program at exactly this point, out of boredom and doubt rather than fear.
A bad stretch with Dark Gold hides first. The account curve stays flat or climbs, because nothing has been closed. Then you open the terminal and equity has separated from the account total, and keeps separating as additional positions open at larger size. From the inside, a normal waiting period and a genuinely unrecoverable one look identical. There is no signal that distinguishes them until the outcome resolves.
Which failure mode drives people to switch off at the wrong moment? Both do, for opposite reasons. Erosion triggers resignation; separation triggers panic. Neither is comfortable, and I do not think either is obviously easier.
If you have never sat through both, that is worth finding out about yourself before you commit capital rather than after. A demo account will not replicate the financial stakes, though it does show you the shape.
Comparing them on the criteria buyers ask about
| Criterion | Prime Scalper | Dark Gold |
| Strategy complexity | Simple in structure: enter, exit, repeat | More moving parts: entry, recovery sizing, group exit |
| Ease of setup | Standard MetaTrader install | Standard install, more inputs to misread |
| Customization | Fewer meaningful levers | Extensive, including recovery parameters |
| Observed performance | +43% over 8 months | +7% over 9 months |
| Loss handling | Closes and books the loss | Holds and increases size |
| Sensitivity to dealing cost | High, positions last minutes | Moderate, positions last hours |
| Funded account compatibility | Plausible, subject to firm rules | Difficult, a martingale’s open loss can breach equity limits |
| Cost | Check the current price on the product page | Check the current figure on the product page |
A few rows deserve more than a cell.
On customization. Dark Gold exposes recovery parameters, and I would leave them alone. Adjusting logic you do not fully understand is how people convert a functioning program into an expensive one. If you must change something, change your position size rather than the internals.
On loss handling. Neither has what I would call genuine risk management in the sense a discretionary trader means it. What they have is a temperament. One closes, one holds. Your actual defense is sizing, and that remains your job rather than the software’s.
On dealing cost. A position held for minutes pays its spread as a large fraction of the expected gain, and gold carries wider, more variable pricing than a major currency pair, so a venue that widens at the wrong moment hurts the fast product more than the slow one. Our spread explainer covers the arithmetic.
On cost. I am not publishing figures, because vendors run promotions and anything written today will be stale within a quarter. What you pay upfront is almost never the expensive part of this decision anyway.
How I would size each one
Position sizing is the actual control you have, so it deserves more than a passing mention. The correct approach differs between these two, which is not obvious from either product page.
For the one that holds losers, work backward from the deepest recorded open exposure rather than from an average, and assume the next episode runs deeper than anything currently on file, because a martingale sets no fixed ceiling on how far a losing basket can float. Doing that arithmetic honestly usually produces an allocation far below the default input, and the gap between those two numbers is the entire exercise.
For the one that closes losers, margin stays comfortable throughout, so the constraint moves elsewhere. Your question becomes how long a grinding stretch of small realized losses you can tolerate before concluding the edge has gone. Eight months of history gives you a feel for the rhythm; it does not hand you a rule.
Three habits worth keeping either way:
Set the allocation once, then leave it. Changing size mid-decline, in either direction, is how a manageable episode becomes an expensive one.
Watch combined exposure rather than individual curves. If anything else runs in the account, the total is the number that ends things.
Write down your abandon condition before you start. Deciding what would make you stop, while calm, beats deciding it at the bottom.
That third one reads like filler. It is the one I would defend hardest, because it converts a panic decision into a plan made in advance.
Which one suits which situation
You are running a funded challenge or any account with a hard equity limit. Prime Scalper, subject to your firm’s specific rules. A martingale’s open loss can breach most challenge limits long before the group recovers, whatever its calm nine-month figure suggests.
You have capital that can genuinely absorb a martingale’s tail, and you understand what you are accepting. Dark Gold, at substantially reduced sizing, with the mechanism and the 91% category worst case both in view rather than just the +7%.
You want the longer evidence trail. Dark Gold has 9 months against 8, which is marginally more, so the evidence behind the two is roughly equal and neither can lean on record length.
You want both. I would push back. Two programs working the same metal during overlapping hours concentrate exposure while feeling like they spread it. If you want a genuine second engine, put it on an unrelated instrument (SEC Office of Investor Education), and our portfolio course covers sizing across several programs.
What these figures cannot tell you
Record lengths are close, eight months against nine, so for once the evidence is roughly equal, though both are short enough that neither has met a genuinely hostile stretch with confidence.
Account counts differ, two against three, and neither is a distribution.
One caveat specific to this metal: several accounts behind our scalper figures run smaller balances, where a fixed dollar loss reads as a steeper percentage. Some decline figures in this category partly reflect account size rather than design quality.
None of this predicts anything. Every number describes what happened, in the past, under specific settings, through 25 August 2026. European regulators found that between 74% and 89% of retail accounts lose money on leveraged products (ESMA product intervention measures). Automation changes how you execute, not the odds you face.
Frequently asked questions
Which is better: Prime Scalper or Dark Gold?
Neither is universally better, though they suit very different accounts. In our tracked accounts Prime Scalper returned +43% over 8 months with a 3% floating loss, a 19% realized drawdown and both accounts profitable, while Dark Gold returned +7% over 9 months with a 9% floating loss, a 4% realized drawdown and all three accounts profitable. Prime Scalper suits hard equity limits and shows its risk openly; Dark Gold is a martingale whose calm figures defer risk rather than remove it. Record lengths are close, so the evidence behind each is roughly equal.
Can you run both on the same account?
Technically yes, though I would advise against it. Both concentrate on one metal, and Dark Gold holds several positions while the other is also open, so your real exposure at any moment is the sum rather than the larger. People routinely size each as though it were operating alone. If you insist, halve both allocations and watch combined exposure rather than two separate curves.
Does a trailing stop help either of these?
Adding one to a multi-position design usually breaks the logic, because the group is meant to close together rather than individually. Single-position programs sometimes accept a trailing exit more gracefully, though the author has generally already tuned the exit for the strategy. My general position: adjust position size, not the mechanics. Changing an exit rule you did not write is a common route to worse outcomes.
How long should you test one before committing capital?
Six to eight weeks on a demo with your own broker is a reasonable floor, and longer for anything entering roughly once per day, since low frequency needs more calendar time to produce a meaningful sample. Demo will not reproduce fills exactly, particularly on metals during volatile sessions. What it does show is whether behavior matches the description, and what the round trip genuinely takes out per position.
What lot size should you start with?
Size from the deepest recorded open loss rather than the average, and assume the next episode runs deeper than anything on record. For a martingale that has floated 9% underwater with no fixed ceiling on how far a losing basket can grow, that means considerably smaller than the default input suggests, particularly if anything else runs alongside. Working backward from a survivable exposure to a position size is the whole exercise, and most people skip it entirely.
How much weight should you give user reviews and Trustpilot ratings?
Less than most buyers do. Public feedback skews toward recent purchasers, who are structurally more likely to be inside a good stretch than a bad one, and dissatisfied users often stop posting rather than update. Ratings also rarely state account size, settings or broker, which are the variables explaining most outcome differences. Treat them as evidence a product exists and functions, not as evidence about risk.
Does gold volatility affect these two differently?
Yes, in opposite directions. Fast in-and-out positions are hurt mainly by widened dealing costs, since the expected gain per position is small. Programs holding groups for hours are hurt mainly by sustained directional moves, because a metal that trends without retracing prevents the group from closing and lets a martingale’s exposure compound. Sharp intraday swings and long one-way runs are different problems, and each design is vulnerable to one of them.
Further reading
- Best robot for gold trading, the wider category ranked
- Prime Scalper alternatives, the full substitute list
- How to trade gold on MetaTrader, practical setup differences
- Why backtest results differ from live trading, the gap that catches buyers
- Best forex VPS for MetaTrader, hosting that affects fills
- Best expert advisors for small accounts, when capital is the constraint
Sources
- World Gold Council, Has gold’s performance structurally changed? and Gold Market Primer: Market size and structure: gold.org
- European Securities and Markets Authority, Product intervention measures on CFDs and binary options: esma.europa.eu
- U.S. Securities and Exchange Commission, Office of Investor Education, Asset Allocation and Diversification: investor.gov
- Algo Trading Space account tracking: 527 funded accounts, 122,837 closed trades, figures current as of 6 August 2026
Risk disclaimer
Trading gold carries substantial risk and is not suitable for everyone. Past outcomes, whether from our accounts or a vendor’s published record, do not indicate future performance. Every figure here describes specific accounts over specific periods under specific settings, and yours will differ. Automated programs can and do lose money. Never commit capital you cannot afford to lose, and consider seeking independent financial advice suited to your circumstances. Nothing on this page is a recommendation to buy any product or open any account.
