Question 1

What is Mercury's AI brain, exactly?

Accepted Answer

A purpose-built quality gate that decides which setups are worth taking. For every setup Mercury detects, the brain estimates the expected net return in R after costs — modelling spread, slippage and swap — and takes the trade only when that expected net R is positive. The model is a LightGBM + CatBoost gradient-boosted ensemble, built from scratch (not a chatbot or LLM wrapper), scoring a 70+ feature snapshot (news proximity, volatility regime, cross-market momentum, session phase, weekly-range position and more), trained on triple-barrier outcome labels with uniqueness-weighted sampling so overlapping setups aren't double-counted. It is tuned per instrument, because gold behaves nothing like EURUSD. The current validation dataset contains 89,258 historical setups across the seven supported instruments.

Question 2

How is the brain validated — what stops it overfitting?

Accepted Answer

Mercury publishes only aggregate, hypothetical backtest evidence — never broker, demo-account or customer-account performance. The current validation is fill-referenced, cost-aware and built from five expanding walk-forward folds. It grades same-bar stop/target ties conservatively as stop-first, then applies one-position-per-symbol, 3R research-day and 1.5R cluster guards. The guarded out-of-sample replay contains 12,718 trades with +0.235R average net expectancy, a 1.67 profit factor and 23.4R maximum drawdown. Block bootstrap, split-half, matched-context, Deflated Sharpe and Probability of Backtest Overfitting checks all passed. Model promotion is explicit and reversible; monitoring cannot overwrite production automatically. These are simulated results, not a forecast, and they do not guarantee future profits.

Question 3

Which ICT/SMC entry types does Mercury trade?

Accepted Answer

Mercury automates the core Smart Money Concepts toolkit: order blocks, fair-value gaps and liquidity sweeps, each confirmed by a market-structure shift and tuned per instrument. Setups are detected on the M5 timeframe with higher-timeframe bias confirmation, then every one passes through the AI quality gate before it can trade. Rather than firing every pattern on every pair, Mercury runs a curated set of high-probability setups per instrument — the brain learns which configurations actually carry edge on each market and rejects the rest. The detection logic follows ICT canon closely: true ATR-based displacement, wick-anchored stops, midpoint limit entries, first-touch enforcement, inversion-FVG handling and a max-age window — details retail bots routinely get wrong.

Question 4

Do I need to keep my Expert Advisor up to date?

Accepted Answer

Yes — always run the latest EA, which you download from the dashboard (current version v3.14.0). Updates ship security and reliability improvements, including authenticated trade-close reporting and a protected bridge so only your account can stream data into the engine. To update: download the latest EA, recompile it in MetaEditor, keep your per-account bridge token in the EA input and re-attach it to the chart. The admin health page identifies outdated or silent installations.

Question 5

How does Mercury handle high-impact news?

Accepted Answer

Mercury auto-pauses entries on tier-1 macro events — NFP, FOMC, CPI, ECB, BoE — from 15 minutes before to 30 minutes after the release, on every affected symbol. That window reflects the empirical fact that the first 30 minutes of post-event price action is the noisiest part of the day and not where the brain has edge. Open positions run to their preset SL/TP; only new entries pause.

Question 6

Can I run Mercury on a propfirm account AND my own account?

Accepted Answer

Yes. Connect each MT5 account separately; switch between them in one click from the dashboard header. Each account has its own equity, risk rules, and trade journal. Prop accounts route through Mercury's prop-firm-aware rules engine — daily/max drawdown, profit target and news embargo are checked server-side before the EA sends any order, with a kill-switch that aims to stop new entries at 80% of the limit used. Built for FundingPips and WeMasterTrade. Broker symbol aliases auto-resolve (NDX100 → NAS100, US500 → SPX500, etc.) so the bot works on every supported broker without you mapping symbols manually.

Question 7

Where can I see Mercury's results?

Accepted Answer

The public Results page publishes Mercury's latest passing aggregate backtest report: sample size, walk-forward design, net expectancy in R, profit factor, maximum drawdown, portfolio guards and null controls. It does not publish individual trades or any broker, demo-account or customer-account performance. The current conservative portfolio replay contains 12,718 out-of-sample trades at +0.235R average net expectancy, a 1.67 profit factor and 23.4R maximum drawdown. Backtests are hypothetical, can differ materially from live execution and do not guarantee future profits.

Question 8

How is per-trade risk capped?

Accepted Answer

Mercury enforces a hard dollar cap on every trade ($100 by default, adjustable per user from Settings → Per-trade risk cap). The cap is enforced server-side BEFORE the order goes out, then re-enforced on the EA against the broker's actual stops_level-padded SL distance. Even if the broker pads SL outward by 100 points, your realised dollar risk stays at the cap. No silent over-risking.

Question 9

What strategy does Mercury use?

Accepted Answer

Mercury automates Smart Money Concepts (SMC) and ICT: order blocks, fair-value gaps and liquidity sweeps, each confirmed by a market-structure shift and tuned per instrument. Every M5 bar is scanned for these setups, with higher-timeframe bias confirmation, and a curated set of high-probability configurations runs per instrument rather than every pattern on every pair. Take-profit targets aren't arbitrary multiples — they're the next genuine draw on liquidity: prior-day highs and lows, session highs and lows, and equal-high / equal-low clusters, detected automatically per pair.

Question 10

What instruments does it trade?

Accepted Answer

Seven instruments are supported on M5: EURUSD, GBPUSD, USDJPY, XAUUSD (gold), SPX500, NAS100 and GER40 (DAX). Each instrument has its own structural settings because gold, major FX and equity indices behave differently. Public performance reporting is deliberately portfolio-level: Mercury does not publish broker-account results or outdated per-pair research figures that came from exploratory edge-finding runs.

Question 11

How does the AI quality gate work?

Accepted Answer

Every detected setup is scored from structural and market-context features such as higher-timeframe bias, killzone, displacement quality, optimal-trade-entry position, inducement and the model probability. The production gate accepts only the configured high-confidence candidates; costs are deducted before research results are reported. In the current hypothetical walk-forward replay, the AI-gated sample contained 17,440 trades at +0.275R average net expectancy and a 1.78 profit factor. After sequential portfolio guards, the conservative report contains 12,718 trades at +0.235R and a 1.67 profit factor. These figures are aggregate backtests, not live results or a guarantee.

Question 12

Is this a wrapped ChatGPT / Claude / generic AI?

Accepted Answer

No. Mercury's AI is a LightGBM + CatBoost gradient-boosted ensemble we built from scratch, trained on our own labelled price data. It has nothing in common with chatbots or general-purpose language models — it does one thing: estimate each setup's expected net return after costs and gate the trade on it. We deliberately avoid generic "trading AI" wrappers because they're not calibrated to our specific strategy patterns.

Question 13

What are realistic returns?

Accepted Answer

There is no responsible fixed return estimate. The public report is a hypothetical research replay, not an account-growth projection. In the current conservative guarded out-of-sample replay, 12,718 trades produced +0.235R average net expectancy, a 1.67 profit factor, +2,989.6R in aggregate and a 23.4R maximum drawdown under the report's fixed 1R outcome policy and sequential guards. R is risk per trade, not dollars or promised account growth. Spread, slippage, latency, liquidity, broker rules and changing markets can make live results materially worse; backtests do not guarantee future profits.

Question 14

Is win rate enough to judge the backtest?

Accepted Answer

Win rate should never be read alone. Under the current hypothetical 1R validation policy, the conservative guarded portfolio replay reports a 65.1% win rate, +0.235R average net expectancy, a 1.67 profit factor and 23.4R maximum drawdown across 12,718 trades. The relationship between win rate, payoff, costs, drawdown and sample size matters more than one headline percentage, and none of these backtest figures guarantees future profits.

Question 15

Is my money safe?

Accepted Answer

Mercury never sees your broker password or accesses your account directly. We are a bridge — the bot sends trade signals, the Mercury Expert Advisor running locally on your MT5 terminal executes them through MT5's own session. Funds stay with you.

Question 16

What does it cost?

Accepted Answer

See the Pricing page for current plans. You subscribe and pay by card via Whop; there is no free trial. Cancel any time — access continues to the end of the period you've already paid for, with no early-termination fees. Monthly plans are non-refundable; multi-month plans (quarterly, 6-month, yearly) are refundable for the unused time after a one-month minimum is retained. Statutory consumer rights that can't be waived still apply.

Question 17

What payment methods do you accept?

Accepted Answer

Card only, via Whop — Visa, Mastercard, Amex, Apple Pay, Google Pay, and bank/SEPA depending on region. We do not accept crypto/USDT. Whop is our merchant of record: they handle global VAT, fraud screening, and disputes. You'll be redirected to Whop's hosted checkout when you click Pay; typical settlement is under 2 minutes.

Question 18

Do I need a VPS?

Accepted Answer

Recommended but not required. MT5 only fires trades while the terminal is running and connected to your broker. A VPS keeps it online 24/5 even when your laptop is closed. If you trade only when you are at your desk, a VPS is optional.

Question 19

Will it work with prop firms?

Accepted Answer

Yes — provided the firm allows EA / algorithmic trading. We list firms we support on the Prop Firms page, and Mercury has a built-in rules engine that aims to enforce max drawdown, daily drawdown, profit target, news embargo, and more on every signal. See the Prop-firm rules section below.

Question 20

What broker / accounts work?

Accepted Answer

Any MT5 broker (this is a vanilla MT5 EA — no broker integration). Prop firms supported / built for so far: FundingPips and WeMasterTrade. The full live list is on the /prop-firms page. Own-capital brokers supported: OctaFX. You stay in full control of your funds — Mercury only places trades, never withdraws.

Question 21

What's the minimum account size?

Accepted Answer

$1,000+ recommended. Below that, the position sizes round to zero on most brokers and the bot effectively paper-trades.

Question 22

Can I cancel any time?

Accepted Answer

Yes. Cancel from your dashboard. Access continues until the end of the billing period you've already paid for. No early-termination fees.

Question 23

Refund policy?

Accepted Answer

Monthly plans are non-refundable. Multi-month plans (quarterly, 6-month, yearly) refund the unused time after a one-month minimum is retained. You can cancel any time to stop the next charge; access continues until the end of the period you've already paid for. Statutory consumer rights that can't legally be waived in your country still apply. Full text on the Refund policy page.