← The Crew

THE CREW

Version 2.84 · XAUUSD · MetaTrader 5

Eleven session specialists under one warden. Equal-lot recovery. You choose how many run.

Symbol: XAUUSD (Gold) — any broker spelling (XAUUSD, XAUUSD+, XAUUSD-ECN, GOLD.pro) Timeframe: any — the EA runs its own internal clock · Platform: MetaTrader 5 · Account type: hedging

This manual is written for someone who has never run an expert advisor before. If you only read one chapter, read chapter 4.

1. What This Is

THE CREW runs eleven independent session-timed sub-strategies on one gold chart. Each wakes in its own hour window, opens a single position, and — if price moves against it — adds equal-sized layers at a fixed dollar spacing, closing the whole group when price returns to a shared average target.

The number of sub-strategies you enable IS the risk dial. That is unusual and it is the point: you are not tuning an abstract percentage, you are choosing which specialists trade. Some of them are structurally shallow and rarely build a deep recovery; others carry most of the account risk on their own. Section 5 gives you what is currently measured about the difference.

Read this first. This is a grid. It holds positions through adverse moves by design — that is how it earns. You will see floating (unrealised) losses on open trades while a group recovers. Sections 4 and 10 are mandatory reading.

These are backtest figures throughout this manual, not results from live trading. Results differ by broker — the measured reason is in section 4.5. Past performance is not indicative of future results. Only trade with money you can afford to lose.

2. How It Works (Plain Language)

Eleven specialists, different hours. Each sub-strategy trades one narrow window of the day and one direction. Some buy the London pre-open, one sells the morning session, one works the small hours. They spread the work across the day instead of betting on a single idea.

Each one recovers rather than stops out. When a position goes against it, the sub-strategy does not panic-close and does not double down. It adds another position of the same size a measured distance away, which pulls the group's average price toward the market. The whole group closes together once price reaches that shared average target.

Adding layers is the engine, not the emergency. This matters and it is counter-intuitive: a depth cap was tested and made things worse — the group could no longer pull its average toward price, so it sat open far longer and the floating loss grew larger, not smaller. Depth caps are therefore not offered as a risk control on the shipped build, because measured, they are not one.

A shared arbiter stops the specialists from crowding. Three of the eleven trade the same hour in the same direction. When they compete for the same moment, a fixed priority decides which one takes it. This makes the EA behave consistently; see section 8 for what it does and does not buy you.

Flat lots inside a group. Always. Every layer of a group is the same size as the layer that opened it. No martingale, no doubling, no escalation while a recovery is running — risk inside a group grows in a straight line. Between groups the size follows the account, so the strategy grows with the balance (section 5b).

A hard account breaker. If floating drawdown reaches your configured percentage of the equity peak, the EA flattens everything and stands down until the next day. An optional confirmation delay (InpBreakerConfirmSec, default: fire on the first tick) lets you require the breach to persist before it acts.

Never refuses a small account. If the balance cannot carry the chosen roster at the chosen risk tier, the shipped build narrows the roster to what the account can carry rather than refusing to trade or silently trading a different configuration than requested.

3. Quick Start

1. Attach to a gold chart (any timeframe) on a hedging account.
2. Load a preset from presets/ — THE_CREW_BAL6_CONSERVATIVE reproduces the shipped defaults and is the recommended starting point.
3. Pick your Roster (section 5) — this is the main risk decision.
4. Pick your Risk tier (section 5b). Conservative is the shipped default and budgets 10% of the balance for floating drawdown; Balanced budgets 15% and grows the lot faster as the account grows — read section 4.5c before choosing it.
5. Check the startup line the EA prints in the journal. If your server clock is not GMT+2/+3, set InpBrokerHourShift — it is the one setting that silently removes the edge if it is wrong (section 7.2), and it must be right before you fund anything.

That is the whole setup. Everything else is already the validated configuration.

The presets

Every preset is emitted complete — all twelve inputs, explicitly — by the build tooling from the .mq5 defaults. A hand-edited or partial .set silently inherits whatever the build happens to default to, which is how you end up trading a configuration nobody validated. Do not edit them by hand.

4. Where the Risk Really Lives

1. It holds floating drawdown on purpose. While a group recovers you will see red on open trades. That is the mechanism working. If you cannot leave that alone, this is not the right strategy for you.

2. There is no stop loss. The only brake is the account-level breaker, and when it fires it realises the loss it was holding. It is a last resort, not a routine safety net.

3. Your account balance decides how the floating loss feels. The same trades produce a very different experience at different balances, because a fixed dollar amount of floating loss is a smaller share of a larger account. Section 5b hands the arithmetic to the EA, which is what makes the position size grow with the account.

4.5c Three live feeds, one tier change (2026-09-06)

Backtest, MetaTrader 5 Strategy Tester, real ticks, 100% history quality, the same binary, the same dates (2025-01-02 to 2026-06-30):

Every tier was positive on every feed with 100% positive months, except the Balanced tier's one flatten on Pepperstone. The same February 2025 basket floated roughly twice as deep on that feed as on Fusion from the identical start date: your broker's tick path decides the depth, not the code. The breaker did its job (the account survived a deep dip and finished positive), but a default tier that can reach its own hard flatten on one live feed is not a default. Balanced remains selectable.

A higher risk tier scales the floating drawdown by roughly the same factor as the profit.

4. Tested in a gold bull market. The validation windows referenced in this manual (2022 through 2026) ran gold from roughly 1,800 to 4,000. There is no bear-market evidence for this strategy. A sustained downtrend is the scenario a no-stop grid is least equipped for, and it has not been measured.

5. Your broker changes the answer. On identical code the worst floating exposure at a fixed 0.01 lot varies by broker feed — backtest, MetaTrader 5 Strategy Tester, real ticks, measured across several broker tick archives, 2022-2026: roughly 140 units of account currency per 0.01 lot on the more forgiving feed measured, and roughly 342 on the least forgiving, because spread decides which entries fire at all. Validate on the broker you will actually trade, not on a convenience feed. The table below turns that cross-feed worst case into a share of balance, at a flat 0.01 lot (before the Risk tier auto-sizing in section 5b changes the lot):

This is safety information, not a performance figure: it is simple division of the measured worst-case floating exposure by a hypothetical balance, at the smallest lot the EA can trade. Your own Risk tier and broker profile change the actual number — the EA prints its own measured figure for your broker and roster in the journal (InpRiskReport, on by default) before you rely on this table.

6. News shocks are the human's job. No statistical model pre-empts a CPI or geopolitical spike. Consider flattening manually before scheduled high-impact events if you are running a wide roster.

5. Performance and Account Sizing — the important section

Honesty note. The figures below are backtest results, MetaTrader 5 Strategy Tester, real ticks, 100% history quality, XAUUSD, 2025-07-01 to 2026-09-07, 10,000 starting balance, compiled defaults (Roster: Balanced 6, Risk tier: Conservative — the shipped default), measured on build 2.83. The shipped 2.84 build was verified behaviour-identical to 2.83 by a trade-for-trade parity test on the same feed and window (identical deal count, profit matched to the cent). Past performance does not guarantee future results. There is no live track record.

The same configuration takes near-identical trades on every broker and earns different money on each: results differ by broker, and OANDA is shown only as a spread-survival check, not as a broker recommendation or a headline.

Widening the roster (Core 4 → Balanced 6 → Wide 8 → Full 9 → Balanced 8) is expected to raise both the return and the floating drawdown together, roughly in proportion. The exact per-roster dollar figures for this binary version are being re-measured; rather than quote the pre-2.63 table (which does not describe the current risk gates and is withdrawn as a current claim), this manual directs you to the number that is always current: the EA's own end-of-run risk report and its start-of-run journal line, which print the measured worst floating drawdown for your broker, roster and risk tier every time you run it. Read that line before choosing a roster wider than Balanced 6, and before funding an account.

Balanced 8 needs an extra caution. Its measured floating drawdown depends on a self-calibrating risk gate that has a warm-up period (section 7.2): a fresh account starts closer to the gate's ungated behaviour than a long backtest shows. Until the shipped binary's Balanced 8 figure is re-measured under warm-start, size it as though its floating drawdown were at least as large as Full 9's.

5b. Risk Tier — how the EA turns section 5 into a lot size

You do not have to size the lot by hand, and you should not be stuck at 0.01 forever while the account grows. That is what the Risk tier does.

Why this works, exactly. This engine is linear in volume: flat lots inside a group, no stop loss, no escalation during a recovery. Trade for trade, doubling the lot doubles the profit and doubles the floating drawdown. Nothing else changes — not the entries, not the spacing, not the depth. So a risk setting can be stated as one honest number: the share of your balance you are willing to see floating. The EA divides that budget by the roster's own measured floating drawdown at 0.01 lots and sizes accordingly:

lot = 0.01 x (balance x budget% / 100) / (worst floating drawdown at 0.01 lots, measured
      for your broker profile and roster, printed in the journal)

Three things to know before you rely on it.

1. The lot is locked when a group opens. Every layer of that group uses the size the first position used, even if the balance moves while the recovery runs. It has to be — equal-lot averaging is what pulls the group's average price toward the market, and that is the only way a group closes. Only new groups pick up the new size.
2. The divisor is measured on the broker profile you select, not automatically on yours. Section 4.5 is not a footnote: worst floating exposure varies meaningfully across feeds. InpBrokerProfile picks which measured table the EA divides by — Vantage (the default) or OANDA (which carries a +1h clock). Backtest your own broker before you trust either one.
3. A Custom roster has no measured number, so the EA assumes the widest roster it knows and sizes small. It says so in the journal.

If the account is too small to reach the requested budget, the shipped build narrows the roster to what the balance can carry rather than under-sizing silently or refusing to trade — read the journal line it prints when this happens.

Margin, not floating drawdown, can be the binding constraint on a low-leverage account. A free-margin guard fixed at 20% blocks new groups when free margin runs low. It never blocks a layer — blocking layers is a depth cap, and section 2 explains why those make things worse. The guard is not a setting: it is baked in, because the value it should take is not a preference.

6. Testing a Small Account on a Large Demo — InpVirtualBalance

Not on the shipped facade. The product build exposes twelve inputs and InpVirtualBalance is not one of them; it is fixed at 0 (off). The mechanism is still in the source and appears when the file is compiled with RESEARCH_BUILD defined. On the shipped binary, fund the demo at the balance you actually intend to trade — that is the only configuration the facade can express, and it measures the same thing more honestly. This section documents the mechanism for anyone building from source.

Brokers rarely let you fund a demo with exactly your intended balance, and a large demo will happily size positions a small account could never carry. InpVirtualBalance rebases the whole risk arithmetic onto a fictional balance:

virtual_balance = InpVirtualBalance + (real_balance - real_balance_at_start)
virtual_equity  = InpVirtualBalance + (real_equity  - real_equity_at_start)

• Position sizing uses the virtual balance, so the demo trades the lots your intended account would actually trade. (Under Fixed lot the size is your fixed lot either way.)
• The account breaker (InpMaxAccountDDPct) and the equity-peak tracking measure against the virtual account.
• The end-of-run journal line reports worst floating drawdown in money and as a percentage of the virtual balance.
• 0 disables it entirely.

Use it like this: fund the demo with whatever the broker gives you, set InpVirtualBalance to your intended balance, and sizing, the breaker and the risk report all behave as though the account were that size. The journal will print a line similar in shape to:

THE CREW v2.84 | XAUUSD | roster=Balanced 6 (6 specialists) | breaker=30.0% | spread<=1.00
THE CREW SIZING: budget 10.0% of <virtual balance> against a measured <X> float per 0.01 lot
        -> <lot> lots. Lots follow the balance.
THE CREW: VIRTUAL BALANCE <virtual balance> (real equity <real equity>) - sizing, the breaker and
        the risk report all measure against the virtual account.
THE CREW RISK: worst floating drawdown <-X> = <Y>% of the VIRTUAL balance <virtual balance>

<X> is the figure measured for your own broker profile and roster — read it from your own journal rather than assume a number from this manual, since it is exactly the quantity under re-measurement (section 5). That is the only honest way to see how the strategy would feel on the account you intend to trade.

7. Settings Reference

The shipped build has twelve inputs. Everything else — the strategy toggles, the gate constants, the ladder geometry, the calendar filters, the arbiter and the research switches — is fixed at the value it was validated on. That is deliberate: the parts interlock, and moving one without the others produces a machine that was never tested. A knob you cannot evaluate is a way to lose money, not a feature.

(The full surface still exists in the same source file. Uncomment #define RESEARCH_BUILD near the top of the .mq5 and recompile and every fixed value becomes an input again — one source of truth, so a research finding and a shipped build can never disagree about what the engine does. Research .set files are not loadable by the shipped binary, and vice versa.)

1. The council — how many specialists trade

The eleven, and what each one is: Nadia (long, 22:00), Amara (long, 03:00), Nerissa (long, 22:00), Evelyn (long, 19:00), Selene (short, 00:00), Thea (short, 23:00), Imogen (long, 08:00-10:00), Clara (long, 06:00-11:00), Juno (long, 10:00-13:00), Nova (long, 22:00), Cassandra (short, 08:00-09:00). Selene and Thea appear in no shipped roster and carry no measured numbers under the current gates. Individual on/off switches exist only in the research build; the shipped facade selects a roster, not a specialist.

2. Your broker — set this once, it matters

A wrong clock does not degrade the edge visibly — it removes it silently, because the specialists are calibrated to market moments and an hour mask only names a moment if the clock agrees. The response is a cliff, not a slope, and there is no safe direction to err in. Check the startup line the EA prints in the journal before you fund anything.

3. Risk — how much of the account is committed (section 5b)

4. Protection

Fixed, and not offered as settings: the free-margin guard (20%, blocks new groups only), the per-direction book exposure cap, the per-specialist entries-per-day limit, the Friday-night cutoff, the scoped arbiter, the one-minute evaluation clock, the entry filter, and every ladder constant (spacing, target, maximum depth) — each specialist keeps its own. None of these is a preference; each is a measurement.

Settings we deliberately do NOT offer as risk controls

Both were tested and both increase risk, and neither is reachable in the shipped build:

• Depth cap. Capping the deepest specialist's layers moved its worst floating drawdown substantially deeper and stretched one group's hold time from hours to weeks. Adding layers is what lets a group close.
• Floating-loss stop. Closing a group once it is underwater by a fixed amount converts recoverable groups into realised losses. Measured on one specialist, a position that would have finished ahead finished behind instead.

The one place the EA will refuse to act is the free-margin guard, fixed at 20% — and it refuses to open a new group, never to add a layer to an existing one, precisely because of the first bullet. A third rejected control is ATR-denominated spacing: widening the ladder in a volatile regime starves the grid of the layers it needs to pull its average toward price, and measured it cost a large share of net while making results more variable across regimes, not less. For a recovery grid, high volatility calls for more layers, not wider ones.

8. The Arbiter — and what it does not do

Three specialists trade the same hour in the same direction. The arbiter is fixed at Scoped: only the highest-priority one takes a contested moment, scoped so that specialists which never share an hour never interfere. It is not a setting in the shipped build.

It is a consistency feature, not a safety feature. Measured across every arbiter setting, the worst equity drawdown was the same — because the deepest drawdown belongs to a specialist the arbiter never suppresses. It costs a small share of net. It ships on because it makes behaviour predictable, not because it reduces risk. If you want less risk, use a narrower roster or a lower risk tier.

9. What to Expect Day to Day

• Most groups close quickly, typically within minutes to a few hours.
• Most groups never add a single layer.
• A small share of groups go several layers deep. Those are the ones that matter for sizing.
• Deep groups usually resolve within a few hours; occasionally one runs for days.
• Long flat stretches are normal — most sub-strategies trade only in their own hour window.

10. Troubleshooting

11. Risk Disclosure

Trading leveraged instruments carries substantial risk of loss. THE CREW is a grid strategy without a stop loss: it holds losing positions through adverse moves and recovers them by adding equal-sized layers. This exposes the account to open-ended floating loss during a sustained adverse trend.

All performance figures in this manual are backtest results on historical data, produced in the MetaTrader 5 Strategy Tester on real-tick history over a period in which gold trended strongly upward. They are not results from live trading and not a prediction. No bear-market evidence exists for this strategy. Broker conditions materially change the outcome. Never trade money you cannot afford to lose, and never run a grid on an account you cannot leave alone through a recovery.

Is it waiting, or is it broken? — the regime monitor

This product earns in a particular market state and goes quiet outside it. That is disclosed above, but a disclosure without a number is not usable: meeting a flat stretch, you cannot tell whether the machine is waiting for its conditions or whether its edge has changed. This section gives you the one measurement that separates those, taken from this product's own record.

The state to watch is the daily Average True Range of the traded instrument, expressed as a percentile of its own trailing 250 days — not a fixed number of dollars or pips, which goes stale as the instrument reprices. Most charting packages will plot ATR; the percentile is simply where today's ATR sits within the last year of its own values.

Not yet measured on this product. The tripwire requires a real-tick record of at least 36 months so the good-state sample can bound anything; the validation runs behind this manual are shorter than that. Rather than print a number from a sample too thin to support it, this section states the method and leaves the threshold open.

What to do in the meantime. Track the daily ATR percentile as described above alongside your monthly result. A flat stretch while the percentile is low is the behaviour this manual already describes. A flat stretch while the percentile is high is the case worth reporting, because that is the combination that would indicate a change rather than a wait.

What this cannot tell you. While the market is outside the good state, a temporary quiet period and a permanent change in behaviour look identical. The tripwire is the earliest point at which they separate, not an early warning. Nothing here is a guarantee; it is the historical extreme of a backtested record, and being beyond it is evidence rather than proof.