Gold EA for Prop Firms: XAUUSD Checks Before Use

A gold EA can automate instructions on a supported XAUUSD account, but the label does not show that it can pass a challenge. Assess the particular file, settings, account and evidence. Software permission, correct execution and a profitable strategy are separate checks.

What does the gold EA actually trade?

Write down the provider, programme, phase, platform, account currency and exact symbol, including any suffix. Spot-style XAUUSD CFDs and exchange-listed gold futures are different contracts. Topstep describes its programme as futures-only; its gold contracts are not an MT4/MT5 XAUUSD account.

Save the symbol specification with the EA version and settings. MetaQuotes documents contract size, tick size/value, volume limits and volume step separately. A number labelled “pip” is unusable until the developer defines its price increment and cash value. A wider stop does not by itself require more account capital: position size, margin and contract restrictions also matter.

Calculate XAUUSD exposure in cash

Fictional worksheet, not broker terms or a risk recommendation: a USD account, a linear contract of 100 ounces per lot, an intended buy fill of 2,011 and an intended stop fill of 2,007 give a $4 price distance. One lot has $400 of planned price loss. For a chosen $100 budget with $4 reserved for charges, $96 / $400 = 0.24 lot, if that volume and step are supported. The price component is $96; actual slippage or costs can exceed the chosen reserve. If the entry and exit are executable fill prices, do not add the same spread again.

Use the actual contract calculation for other account currencies or instruments. Round down to the allowed step and check margin and existing positions. If the minimum size exceeds the budget, the result is no acceptable order at that stop, not permission to round up or move the stop arbitrarily.

Do not confuse a daily limit with a safe trade budget

For a fictional $500 daily allowance, three $200 losses total $600: 120% of the allowance, not 60%. Two total $400 before any other costs or open loss. The real provider may include floating P/L, commissions, swaps and a particular reset timezone. Track the whole account, including manual orders and other EAs; a pip-loss counter alone is insufficient.

Turn strategy labels into testable rules

• Trend: specify the bar timeframe, completed-bar signal, entry order, stop and exit. Define whether an exit closes a long or opens an independent short.
• Mean reversion: define the reference level, invalidation and maximum exposure. A tighter stop does not establish lower risk or a winning expectancy.
• Breakout: define the range, order activation and handling of rejected, partially filled or duplicate orders.
• Grid or adding to positions: document aggregate exposure and the adverse path. Check the actual programme policy rather than assuming a universal permission or ban.

These are research hypotheses, not strategies shown here to pass. Compare them after costs without retuning repeatedly on the period used to judge them.

Test loss controls and failure behaviour

Prepare a dry-run log with expected and observed outcomes: a restart during an open trade; a disconnected terminal; an invalid volume or unavailable symbol; a rejected close; an existing manual position; a daily reset while a loss is open; and simultaneous signals from two charts. Check that the EA does not duplicate orders after reconnecting and that a pause setting really distinguishes new entries from management of existing positions.

Use the actual account’s loss reference and reset clock. An EA requesting a close at a threshold cannot guarantee the fill or prevent every breach. Do not remove protective orders merely to suppress an alert. Record outstanding orders and reconcile the account after a failed request before deciding whether another request is appropriate.

Read programme permission before choosing a preset

Sources checked on 16 October 2026: FTMO’s CFD strategy FAQ permits EAs subject to its other conditions, including forbidden practices, server activity and allocation issues for identical third-party strategies. That is not approval of a named gold EA.

FundedNext’s CFD EA policy is materially different: eligible MT4/MT5 accounts must be below $50,000 and need the paid EA add-on; accounts from $50,000 must be traded manually and prohibit automated tools even when they only modify SL, TP or lot size. cTrader and Match-Trader automation is prohibited. Distinct-strategy conditions apply, and EAs designed specifically to pass prop-firm challenges are prohibited. A low-risk setting does not override those restrictions.

News permission is another check. FTMO’s news FAQ restricts execution on affected instruments from two minutes before to two minutes after selected releases, including SL, TP and pending-order execution; Challenge, Verification and Swing accounts are exceptions. XAUUSD is affected by the listed USD events. Read the precise current account terms rather than transplanting a single rule to every phase or provider.

What evidence should a developer provide?

Request the exact file and input settings, symbol/server, test dates, data provenance, execution mode, costs, timezone and complete report. MT5’s own documentation describes different execution models and historical floating spreads. “All backtests use perfect candle-open fills” is incorrect; a “99%” badge or a fixed number of years also does not prove robustness.

Keep parameter-selection data separate from the period used to assess the final choice. Review losses, exposure and failures as well as aggregate profit. A forward observation checks behaviour under its observed conditions; no universal two-week or two-month duration establishes a pass probability. A profitable test in several windows is evidence from those tests, not proof of a future edge.

Distinguish the JPTC products

EA Hub is downloadable software for a customer’s supported MT4/MT5 terminal. Check the actual file’s documented features and effective inputs. The separately hosted Algo needs no customer installation or VPS. Neither a product label nor a preset establishes account permission. Company-published broker-statement figures concern that reported Algo account and period; they are not an independent audit or a study of another EA’s challenge pass rate.

The audit outcome can be “specified and tested”, “missing evidence” or “not permitted for this account”. It should not become a promised return, a required purchase or a deadline for passing.