Today, this is the robot we ship

Goldy is in production and ships under the key that covers the Tuurt Trading package: one licence, not one per robot. Cerberus was withdrawn from sale in August 2026 — we measured it on our own account and it did not produce results that justify it — so if you write to us today, Goldy is the robot we set you up with. Anyone who already held a licence covering Cerberus keeps it.

Goldy is not a stripped-down build of anything: it carries the full risk governor — per-basket stop and daily-loss lock — set at the point that came out of measurement. Further down is how to sign up.

Read this before you scroll further

Goldy trades leveraged CFDs on gold and averages down with a ladder that increases the lot size leg by leg. It can cost you your entire capital. It is not financial advice, it is not a promise of returns, and no past result guarantees a future one.

On this page we publish the method we measure with, and also what went wrong: the day both guards fired and it closed in the red, the win rate that does not mean what it looks like, and the worst floating loss of our real accounts, right next to the net. If you are looking for a page that only shows you the pretty curve, this is not it.

How it works

It does not look for a signal. It leaves the trap set.

Most gold robots try to predict where price is going. Goldy does not try: it places two pending stop orders on the same chart, one above price and one below, and lets the actual move decide which one fires.

01 — THE TRAP

Two pending orders

On XAUUSDm, two pending stop orders sit symmetrically around price: a buy above, a sell below. Neither is a prediction; they're the bet that, sooner or later, price breaks in one direction with enough force to trigger one of the two.

The market picks the side. There is no trend indicator or bias filter before the trap is set. The order that fires is the correct answer to where price moved, not a prior opinion about where it was going to move.
The distance is fixed, not guessed. Each order sits 100 pips from price. Gold quotes with three decimals at this broker, so a pip is $0.01 and the trap really opens $1.00 per ounce on each side.
The order that doesn't fire gets cancelled. The moment one triggers, the other stops making sense and is pulled: only one position stays alive per cycle, not both at once.
It re-arms with price still moving. Once a cycle closes, the pair of orders gets placed again around the new price. The robot doesn't wait for the market to sit still: it operates while it moves.

02 — THE GOVERNOR

Risk, kept apart from the signal

The robot Goldy replicates already exists and has traded this way for a while: the two-pending-order trap isn't a new idea. What Goldy adds on top is what that original doesn't have: a hard limit on how much it can lose before something — or someone — stops it.

Per-basket stop. If the open position racks up a floating loss past a set percentage of capital (InpBasketStopPct), the governor closes it and waits out a cooldown before setting the trap again.
Daily-loss lock. If the account drops the set percentage (InpDailyLossPct) from the day's reference, the robot flattens everything and refuses to arm again until GMT midnight or until you release it by hand.
The ladder's ceiling comes from margin, not from faith. The maximum number of legs (InpMaxGridSteps) is a cap, and on top of that the robot recomputes it against live equity (InpMarginBufferPct) before adding each one.
Both guards have already fired in production. They are not a design promise: on its first day trading, both fired, in order and with nobody intervening, and the day closed in the red.

Seeing it move

What happens when the market goes against it.

The sequence that matters most before you sign up. The good cycle is boring: one side fires, price carries on a little, the position closes green. This is the other one — the one that decides whether the account survives.

One side fires and the other order is cancelled

Price reaches one of the two pending orders. That one enters, the opposite one is pulled, and a single position stays alive with the direction the market chose.

Price turns around and the ladder adds legs

Here is the real risk, and we'll say it plainly: the ladder averages down and each new leg is larger than the last. Exposure grows while price moves against you, and the cycle's profit is capped while the potential loss is not.

The ladder stops where margin says it stops

The leg cap is not just a number in the configuration: before every addition the robot checks it still has enough free margin. We put that in because we measured the original reaching 12 legs against a computed ceiling of 9 — it survived because price came back, not because it was protected.

The per-basket stop cuts the whole position

If the floating loss passes the limit, the entire group closes at once and the robot waits out a cooldown. This is the brake that turns an open, open-ended loss into a closed, known one.

The daily lock calls the day over

If the day's drawdown reaches the limit, it flattens whatever is open and stops arming the trap. It does not retry that session: it waits for GMT midnight or for you to release it. A robot that re-enters on the same bad day is how one bad day becomes an empty account.

Live accounts

What the real accounts have measured so far.

Goldy runs today on 8 accounts of ours with real money. None of these figures is typed into this page by hand: they are read from those accounts every time you load it. Next to the net sits the worst floating loss, which is the half that hurts.

Account	Net measured	Worst floating	Max legs	Days measured	Since
TUURT TRADING - T1	5,141.90 ¢	-3,379.40 ¢	14	7	2026-08-24
TUURT TRADING - T2	4,016.60 ¢	-5,813.90 ¢	15	7	2026-08-24
TUURT TRADING - T3	4,147.20 ¢	-7,764.00 ¢	15	7	2026-08-24
TUURT TRADING - T4	5,014.10 ¢	-4,401.30 ¢	14	7	2026-08-24
TUURT TRADING - T5	6,596.40 ¢	-25,014.90 ¢	17	7	2026-08-24
TUURT TRADING - T6	4,640.90 ¢	-10,921.40 ¢	16	7	2026-08-24
TUURT TRADING - T7	3,764.70 ¢	-3,989.70 ¢	14	7	2026-08-24
TUURT TRADING - T8	4,539.20 ¢	-3,640.10 ¢	14	8	2026-08-23

This is a short window, and it is not a track record yet

The days measured in the table are all there is: the measurement started recently and is still running. Over a window this short, a good number and a bad one look far too much like luck. Read it as a measurement that has just begun, not as a track record.

The worst floating loss is sampled every 60 seconds, so it is a minimum: a spike shorter than that interval does not show up. It is the lowest the open position got to, and that is why it sits next to the net instead of hidden underneath it.

The figures ending in ¢ come from a cent account: its balance is denominated in cents, so the number is exactly the one MetaTrader shows. One dollar is 100 ¢.

What we measure

The numbers, with their fine print attached.

You will not find a hand-typed returns figure on this page. Not because we don't have any, but because a figure typed into a landing page ages in silence and nobody can check it: the ones above are read from the real accounts every time the page loads. The full detail lives on a separate page that reads those same accounts, and there we publish the investor credentials so you can verify it yourself in your own MetaTrader.

See the day by day and the investor password

How it is measured, so you know what you are looking at

Net is the sum of real closes, nothing else. No open trade counts as profit that has not been realised yet.

Deposits and withdrawals go separately and never inside the profit. Money that came in is not money the robot made. The robot itself keeps that account: it sums the day's cash operations so they cannot pollute its daily reference. We learned that the hard way — a top-up of the test account read as an 80.6% drop and locked the day with the robot never having opened a single position.

Worst floating is sampled every 60 seconds, so it is a minimum. A spike shorter than that interval does not show up. Read that column as "at least this much", not as "exactly this".

This robot's win rate does not mean what it looks like

Measured on our own accounts, Goldy closes green around 97% of the time. It is the number anyone would print in large type, and it is exactly the one you should not be looking at.

That 97% comes from not cutting losers. A system that only closes when it is in profit and leaves the rest open can show you any win rate you like. The original Goldy replicates racked up 1,085 winners against 2 losers over 204 hours, and that record did not come from avoiding the loss: it came from deferring it.

The worst unrealised loss we have measured has run larger than the accumulated net profit. Put another way: there have been moments when what was open in the red exceeded everything the robot had made up to that point. That is why both guards exist, and why the track record page publishes worst floating next to net.

Judge it by net USD and by worst floating. Never by the win rate.

Requirements

What it needs to run.

The whole list is short and holds no surprises. If you are missing any of it, tell us before we set you up and we'll talk it through.

MetaTrader 5, not MetaTrader 4

Goldy is written in MQL5 and runs on the MT5 Windows terminal. There is no MT4 build and none is planned: Cerberus is the MT4 project, and it is no longer for sale.

An account carrying the gold symbol

It trades XAUUSDm. The exact symbol name varies by broker, and your account's minimum lot size determines how many ladder legs fit before margin takes over. Tell us your broker and we'll check it with you.

The licence key, pasted into the terminal

The key arrives by email and goes into the robot's parameters. It revalidates against our server every few hours, so the terminal needs that address authorised in its options. It is the only manual adjustment in the process.

A VPS, if you're leaving it unattended

In practice it is not optional, and it is worth knowing why: both guards live inside the robot's code. With the terminal off, nobody is measuring the floating loss or the day's drawdown. A VPS is what makes the governor exist 24 hours a day.

How to get started

One email, one key, paste it into the terminal.

Goldy ships under the key that covers the Tuurt Trading package — if you already have a Cerberus key, write to us and we extend it to Goldy. If it's your first time, write with your account number and broker, and the key arrives by email to paste into the terminal.

We don't publish a price on this page: it depends on the broker and your account size, and we'd rather talk it through than give you a number with no context.

trading@tuurt.com Or on Telegram → See the Tuurt Trading package →

Hard questions

The ones you'd ask with ten minutes in front of the engineer.

It really needs no signal at all?

Really. There is no trend indicator, no bias filter, no model with an opinion about gold before the trap is set. The two stop orders go in and it waits.

That is the part of the original robot we thought was worth keeping: the market picks the side, and an entry that predicts nothing cannot be wrong about the prediction. What it can do is enter a move that immediately reverses — and that is where everything else on this page comes in.

What happens at the weekend, with the market closed?

Gold closes Friday at 21:00 UTC and does not reopen until Sunday at 22:00. That is 49 hours in which an open position hangs: nobody can close it, the guards cannot act, and the price it reopens at can be a long way from the last one.

That reopening gap is a real risk and no setting covers it. In this gold family we have measured a reopen $35 above Friday's close. If you are going to leave the robot running over the weekend, do it knowing that.

There is also a daily broker pause between 21:00 and 22:00 UTC in which the server rejects any order. The robot crosses it without trading; the price spikes that appear in that hour, with the market barely quoting, are not real moves.

What exactly do the two guards do, and how do they differ?

The per-basket stop watches the open position. If its floating loss passes a percentage of capital, it closes the whole group and waits out a cooldown before setting the trap again. It can act several times a day if it has to.

The daily-loss lock watches the account. If the day's accumulated drawdown reaches the limit, it flattens whatever is open and refuses to arm again until GMT midnight or until you release it by hand. It acts once and calls the day over.

The first limits how much one cycle hurts; the second limits how many bad cycles you can chain. On Goldy's first day trading, both fired, in that order.

If it wins 97% of the time, why isn't that printed in large type?

Because in this kind of system the win rate is a knob, not a virtue. Closing only when you are green and leaving open whatever is red produces any win rate you like — until the day what is open does not come back.

We measured it on the robot Goldy replicates: 1,085 winners against 2 losers over 204 hours. That record did not come from avoiding the loss, it came from deferring it. And on our own accounts, the worst unrealised loss has run larger than the accumulated net profit.

Look at net USD and worst floating. Both are on the track record page, with the investor credentials so you can check them yourself.

Is this a martingale?

Yes, in plain words: the ladder increases the size of each new leg while price moves against it. We are not going to call it anything else.

What separates it from a classic martingale is not the multiplier, it is the ceiling: the number of legs is capped and additionally recomputed against live equity before every addition, and above all of it sits the per-basket stop, which can close the group before the ladder ever reaches its cap. The original we replicate got to 12 legs against a computed ceiling of 9 and survived because price came back. That is precisely what Goldy does not allow.

Even so: averaging down increases exposure while you are losing. The ceiling changes the ending, not the nature of the ride.

What if my VPS goes down with a position open?

With the robot off, nobody is measuring. Both guards are code running in your terminal: with no terminal there is no per-basket stop and no daily lock. And with the ladder active, the position carries no resting stop order on the broker's server covering it in the meantime.

That is the reason we insist on the VPS, and we'd rather say it here than find out together.

Do you guarantee returns?

No. Neither we nor anyone being honest with you. What we do guarantee is that the protection rules do exactly what this page says, that the published results come from our own accounts and can be verified with the investor credentials — including the days that went badly — and that if anything changes, we update it here.

Risk disclosure

Goldy is a high-risk experimental system. Nothing on this page is an invitation to trade, whether with real money or with a demo account of your own, and it is not investment advice. What follows applies to the robot we ship and to our own soak account, where we keep measuring its most aggressive configuration.

It trades a high-risk instrument. XAUUSDm is leveraged gold via CFDs. It can cost you the entire capital in the account.
The payoff is asymmetric and the ladder is a martingale. The profit per cycle is capped; the loss is not. Each new leg is larger than the last and adds exposure while price keeps moving against it.
A high win rate is not an edge. Goldy's — around 97% — comes from not cutting losers, and the worst unrealised loss we have measured has run larger than the accumulated net profit. Judge it by net USD and by worst floating, never by the percentage.
The worst floating we publish is a minimum. It is sampled every 60 seconds: a spike shorter than that interval is never recorded.
It only defends while it is switched on. Both guards are code running in your terminal. With the terminal off, and with the ladder active, there is no protective order waiting on the broker's server.
It does not protect against gaps or against the broker. Gold closes Friday at 21:00 UTC and reopens Sunday at 22:00; an open position crosses 49 hours with nobody able to close it. Slippage, spread widening and requotes are equally outside its control.
The build running on our own soak account is deliberately the most aggressive one possible. Without the brakes the configuration we ship does carry, precisely to find the limit — that's how we know where the edges of what we ship are.
No number from this soak account has been validated with the rigor we require before publishing a configuration, and none of it runs on any customer's real money.
Nothing published here is investment advice or a promise of returns. No past result, measured or projected, guarantees a future one.
Trading leveraged CFDs carries a high risk of losing money rapidly, including the entire deposit. You alone are responsible for what you run in your account.

Shall we talk before you decide?

Write to us with your account size and your broker. We'll tell you whether it makes sense in your case — and if it doesn't, we'll tell you that too.

trading@tuurt.com