Mean Machine AI is an automated Expert Advisor for MetaTrader 5. It reads the market, decides which of its eight engines fits the conditions, places the orders, manages the baskets and closes them — while you are at work, asleep, or nowhere near a screen. Nothing to copy, nothing to time, no hours in front of a chart.
Every engine owns its own magic range, basket, hedge and exit. A hard cap keeps three groups live at once at most, and scalp and HFT can never hold positions at the same time.
Layered entries at an ATR-scaled step with a take-profit ladder. Arms its hedge when basket drawdown crosses its threshold.
The mirror side. Both can be live at once; the anti-hedge magic manages the pair as a single exposure.
Tighter grid, tighter targets, short holds. Blocked automatically while the HFT engine holds a position.
Same rotation on the sell side. Spread is capped as a percentage of grid step, so expensive pairs stand down instead of bleeding.
Fades stretch away from the mean, but only while Bollinger width says the market is ranging rather than expanding.
Triggers on a sudden move, then requires the adverse leg to stall across three M1 bars before each add. Grid widens 25% per adverse add, capped at 3×.
Only fires when D1 ADX clears 32.5 and EMA 50/144 separation clears 0.36%. Donchian entry, Chandelier trail, partial at 2R, exit on regime flip.
Waits for price to return to the fast EMA inside an established H1 trend. Break-even at 0.75R, scale-out at 1.5R, profit trail from there.
The regime classifier runs on the daily chart and gates everything below it. If the market does not qualify, the directional engines do not fire. No trade is a position.
ADX(22) on D1 against a 32.5 floor. Below it the chart is treated as a range no matter how it looks.
Distance between EMA 50 and EMA 144 as a percentage of price. Converging averages mean no conviction.
ATR is measured against a gold reference of roughly 6,000 points a day, then every grid step, target and lot on other symbols is scaled to match.
A squeeze flags a coming expansion. Mean reversion stands down, the spike engine arms.
H1 and M15 direction are checked against the daily regime before a pullback entry is accepted.
Commission is auto-detected from your own closed trades, so a profitable exit is profitable after costs.
The retracement must be at least 1 ATR deep and actually touch the EMA. Shallow drifts are rejected.
Each engine tracks its own peak and drawdown, arms its own hedge, and reports both live in the terminal panel.
Three strategy groups live at once, maximum, plus a global safety stop that flattens everything past your exposure cap.
One .ex5 file, one chart per market. A regime classifier on the daily chart decides which engines are allowed to trade, each engine owns its own magic range, basket, hedge and exit, and every distance it uses — grid step, target, stop, lot — is expressed as a percentage of that symbol's own average daily range rather than a fixed pip count. Drop it on gold, on a JPY cross, or on a broker that quotes five decimals instead of three, and the grid comes out the same shape.
The autopilot reads ATR on the daily chart and derives every pipstep, every take-profit ladder and the spike trigger from it, then recalibrates on each new daily bar so it tracks volatility instead of assuming it. That is why the same file runs on metals, majors and crosses without a settings hunt.
Each engine tracks its own peak and drawdown and arms its own hedge when that drawdown crosses its threshold. A hard cap keeps three strategy groups live at once at most, scalp and HFT can never hold positions simultaneously, and a global safety stop flattens everything past your exposure cap.
This is software you run, not a fund you join. We never hold your capital, never trade on your behalf and never have access to withdraw anything. The account is yours at your own broker, the bot runs on your own machine or VPS, and everything it does you can undo.
No pooled fund, no managed account, no PAMM. You deposit with your broker, in your name, and you are the only person who can withdraw it. If we disappeared tomorrow the money would still be sitting where you put it.
Close a position by hand, move a target, flatten a whole basket — the bot manages what is open, it does not lock you out of your own terminal. Nothing needs our permission and nothing needs to be unlocked first.
Every strategy has its own on/off switch in the inputs — UseBaseStrategy, UseScalpStrategy, UseMeanReversionStrategy, UseHFTStrategy, and Enable_EngineA and Enable_EngineB for trend and pullback. Set one to false and that engine never opens a position. Run all eight, or run one.
The hedges have their own switches too, so you can run an engine with its hedge armed or without it, per engine, rather than accepting one global setting that covers everything.
Turn AutoTrading off, or drag the bot off the chart. There is no notice period inside the software and no way for us to keep it trading after you have said stop.
This one is deliberate and worth knowing: if your subscription lapses, or our licence server is unreachable, the bot stops opening new positions and keeps hedging, trailing and closing whatever is already open. A billing problem must never liquidate a live basket, so it cannot.
On Friday 30 January 2026 gold fell roughly 8% intraday, and by some measures more than 12% from its high — its worst single day since the April 2013 crash, and on some readings the largest one-day drop in close to four decades. Silver fell further still, posting the largest intraday decline in its recorded history.
A one-way move of that size is the exact shape a grid is worst against. Entries fill all the way down at spacing that was set for a normal week, the basket never gets the retrace it needs to resolve, the account runs out of margin before the strategy runs out of levels, and it is the broker that closes the position rather than the bot.
This is not a knock on grid trading. It is the specific failure mode, and any honest vendor should name it before you fund an account.
Because every distance is a percentage of the symbol's own ATR, a volatility spike widens the grid instead of stacking new entries at last week's spacing. The spike engine requires the adverse leg to stall across three M1 bars before each add and widens its own grid 25% per adverse add, capped at 3×. Each engine arms its hedge on its own drawdown rather than waiting for an account-level threshold, and the concurrency cap stops three engines from loading the same move at once.
We ran the bot through that week and recorded it. Those recordings are below — watch the drawdown rather than the equity curve, and watch what the hedges do while the basket is under water.
You can also check it yourself without taking our word for anything: the XAUUSD run on the Backtests tab opens on 27 January 2026, three days before the crash, so that day sits inside the tested window rather than being quietly excluded from it. The balance curve there is the real closed-trade balance from the report.
One event is not a track record, and a strategy that came through January is not thereby proof against the next shock, which will have a different shape. The honest test is your own: run it on a free demo account through a full month, including a bad one, before you fund anything.
Strategy Tester runs and live sessions, recorded end to end — including the parts where the basket is deep under water and the hedge is doing the work.
Recordings are being edited and will appear here shortly.
There is no broker lock-in. The one hard requirement is a hedging account, because the engines hold opposing positions on purpose and a netting account would collapse them into one.
Brokers write gold as XAUUSD, XAUUSD.m, GOLD or XAUUSD-ECN. The licence server normalises all of them, so your broker's suffix will not lock you out of what you paid for.
Commission is auto-detected from your own closed trades, so a target that looks profitable on the chart is profitable after what your broker actually charges you.
Spread is capped as a percentage of grid step. On an expensive pair the scalp engines stand down instead of paying the spread on every rotation.
They offer MT5 hedging accounts with the symbol names and contract sizes the bot is tested on, so there is nothing to adjust after you connect.
Some brokers advertise deposit bonuses. Do not count one toward the recommended balance: bonus credit usually cannot absorb losses or be withdrawn, it can be removed when equity falls, and EU-regulated brokers may not offer bonuses on CFDs to retail clients at all. Fund the balance with your own money, and check which of the broker's companies you are signing up with and that it is allowed to serve clients where you live.
We earn a commission from the broker if you register through our link. It costs you nothing extra and does not change the software price.
Keep it. Check three things: the account is hedging rather than netting, gold spreads are competitive during the London session, and the platform is MT5 rather than MT4. If all three hold you are done — nothing else about the bot changes.
If installing an EA on a VPS, whitelisting a WebRequest address and matching a set file to your balance sounds like someone else's job, it can be. One-time fee, not a subscription, and the only thing we charge for beyond the licence.
Two things. Buy a VPS — around €5–10 a month, and we tell you which one and which location for your broker. Then open an MT5 hedging account — with one of our partner brokers or one you already use.
That is the whole list. You do not install anything and you do not touch a setting.
Everything else: install and configure MT5 on your VPS, install the bot, whitelist the licence address under Tools → Options, load the set file matched to your balance and market, attach it to your charts, confirm the licence panel is green and the engines are enabled, and stay with you until you can see it running yourself.
You keep full control of the account and the VPS throughout — we never ask for your withdrawal password.
This is grid and hedging software. An open basket can carry significant unrealised drawdown before it resolves, and losing an entire account balance is a real possible outcome rather than a disclaimer. Use money you can afford to lose.
Backtests, recordings and demo results describe what already happened. They do not predict what happens next, and we do not promise profit, manage your money, or give investment advice.
Market figures for 30 January 2026 as reported by Finance Magnates and TradingKey. Intraday and close-to-close measures differ; both are cited above.
Every run below is a Strategy Tester result on historical data, not a live account. Read the maximum drawdown column first — that is the number that decides whether you can actually sit through this.
| SYMBOL | PERIOD | NET RETURN | MAX DD | PROFIT FACTOR | TRADES | WIN RATE | SOURCE |
|---|
This one is a real Strategy Tester export from the standard settings, and the curve above is the actual closed-trade balance from that report rather than a drawn shape. Five things decide what it is worth to you.
Every engine was enabled for the whole run. No engine was switched off for the conditions and no period was excluded, so the result includes the trades the wrong engine took on the wrong day as well as the ones that worked.
It is eight months, one symbol, one broker's history. That is not a cycle. A strategy needs to be wrong in several different ways before you know how it fails, and eight months rarely supplies that.
It starts at €500, on purpose. That is half our own recommended minimum of €1,000 per market, so the grid is working with half the room we would want it to have — which is the point of testing it that way. Read the return with the matching caveat though: compounding from a small base produces large percentages that do not repeat at scale, and lot sizing is balance-relative, so the same run on €5,000 does not give you five times the percentage.
Gold in 2026 was exceptional. The tested window opens on 27 January, three days before the historic crash, so the run contains precisely the kind of volatility this bot is built for. A quiet year looks different, and you should expect it to.
The worst equity drawdown was about 30% of the account at that moment, and the largest in cash was €502.67 — more than the original deposit, at a point where the account had grown past it. That is the number to sit with before the return, because it is the one you have to hold through.
The commission input was left at zero in this run. Your broker's round-turn commission comes off the top of these figures. Live, the bot measures commission from your own closed trades instead of assuming it.
And the general caveat still applies: a tester cannot reproduce slippage, requotes, or a broker widening spreads during a release. Run it on a free demo for a month before you fund anything, and if a live month disagrees with the backtest, believe the live month.
Backtests are curve fits until proven otherwise. They use historical spread, they cannot reproduce slippage, requotes or a broker widening spreads during news, and grid strategies flatter themselves more than most because an open basket that eventually resolves looks harmless on a chart.
Run the same settings on a demo account for at least a month before you fund anything. If a live month disagrees with the backtest, believe the live month.
Placeholder data. This row has not been replaced with a real Strategy Tester export yet. The rows marked verified report in the table are real; this one is not, and it must not be published as though it were.
Pick the markets first. The price follows from how many you run, and the recommended balance follows from that.
Each market you pick is one seat — one symbol on one live account. Run them all on a single account or spread them across several, it costs the same. Seats can be freed and moved from your account page.
Each licence runs its own grid with its own drawdown. They do not share margin, so the balance requirement adds up rather than averages out.
The bot needs a hedging account, tight gold spreads and matching symbol names. These four offer MT5 hedging accounts with the symbol names the bot expects.
We earn a commission from the broker if you register through our link. It costs you nothing extra and does not change the software price. Already have an account you like? Skip this step — the bot works on any MT5 hedging account.
.ex5, set files for your account size and the manual. Paste the key into the bot's inputs — each chart claims its seat on its first check-in, so there is no account number to register. Cancel any month: the licence runs to the end of the paid period, then the bot stops opening new trades and keeps managing the open ones.One symbol on one live account. Demo stays free and unlimited.
Two symbol-and-account pairs. The usual starting point.
Three pairs, which is also the concurrency cap on a single account.
Fifteen pairs, spread across as many accounts as you like.
Everyone gets the same .ex5 — you never receive a separate file per market. Install it once, open one chart per symbol, and paste the same licence key into each. Give each chart its own Base_MagicNumber, 100 apart, so the engines never collide.
You never type your account number anywhere. The bot reads it from the terminal and sends it with the check-in, which happens on load and every six hours after. The first check-in from a symbol-and-account pair claims a seat; the same pair checking in again just refreshes it.
One-time setup: allow our address under Tools → Options → Expert Advisors → Allow WebRequest. The manual shows exactly where. Manage your seats any time at your account page.
Grid and spike engines need an uninterrupted connection. A laptop going to sleep mid-basket is the most common cause of a bad month. A low-latency VPS near your broker's server costs about the price of a coffee a week.
See the VPS we useAffiliate link, same disclosure as the brokers.
Trading leveraged products carries a high risk of losing money rapidly. Mean Machine AI uses grid and hedging techniques, which means an open basket can carry significant unrealised drawdown before it resolves. Losing your entire account balance is a real possible outcome.
Past performance, backtests and demo results do not predict future results. Every figure on this site is historical and labelled as such. We do not promise profit, we do not manage your money, and nothing here is investment advice. Use money you can afford to lose, and test on demo first.
Price charts are supplied by TradingView. The regime analysis beside them is illustrative unless marked live.