Multiple trading bots on one account: diversify or multiply?

One bot is running well, so you add a second. Then a third. Sounds like diversification. More strategies, more instruments, less dependent on any one thing.

Usually it is not. Separate trades are not separate risk.

Can you run multiple trading bots on one account?

Yes. MetaTrader 5 runs one Expert Advisor per chart, up to 100 charts per terminal, and a unique magic number keeps each bot's trades identifiable. But separate trades are not separate risk. Every bot draws on the same free margin and the same equity, so margin level, margin call and stop out apply to the account as a whole, never per bot. One bot's drawdown can trigger a stop out that closes another bot's winning position.

So the plumbing is the easy part. One EA per chart, and if you try to attach a second to the same chart, MT5 asks whether it should remove the first. On the MetaQuotes VPS the ceiling is 32 charts.

The risk is the hard part, and that is what the rest of this is about.

What is a magic number and does it really keep my bots apart?

A magic number is a number an EA attaches to every order it places. The MQL5 documentation describes it as the ID of the Expert Advisor that placed the order, designed so that each EA uses its own unique number. That is how a bot finds its own positions among everyone else's.

Two things the docs do not say, and they surprise a lot of people.

MT5 does not enforce one bot per symbol. You can put three bots on EURUSD and the platform is perfectly happy.

And a magic number only works if the code uses it. The function that returns all open positions does not filter by magic number, the developer has to do that. If one of your bots filters only on symbol and not also on magic number, it will start managing its neighbour's positions. Closing them, moving them, modifying them. One badly written EA is enough to make a mess of the whole account.

So a magic number is no more than a label. It makes trades identifiable, but it keeps no money apart.

Why two bots on the same pair wreck each other

This is not about odds. It is about how the platform works.

An MT5 account is either netting or hedging. On a netting account there is only one position per symbol. MetaQuotes puts it plainly: further operations on that instrument only lead to closing, reversing or changing the volume of the existing position. On a hedging account you can hold several positions per symbol, including opposite ones.

Now put two bots on EURUSD on a netting account. Bot A is long. Bot B gets a sell signal. No second position appears. Instead bot B reduces, closes or reverses bot A's position. One bot's stop loss is literally the other bot's entry. Your magic numbers are cosmetic there, because the positions were merged before the label mattered.

At which point you think: fine, hedging account then. That is the other side of the same coin. On a hedging account each position is tracked separately, so your broker is also allowed to charge margin on each one. Some brokers let long and short offset each other, others do not, and it is configured per symbol. Hedging accounts often carry worse swap rates too.

Short version: netting merges your bots and breaks their logic. Hedging leaves their logic intact and eats your margin. There is no account type that solves this. You are picking which of the two failures you prefer.

Can one bot close another bot's winning trade?

Yes, and that is exactly why "separate trades" is misleading.

Your margin level is equity divided by used margin, times a hundred. Look in the MT5 API and every related property sits at account level: margin level, margin call level, stop out level, free margin. There is no per-EA equivalent anywhere. The regulator works the same way: the FCA requires a broker to close out a customer's position once their funds fall to 50% of the margin needed to maintain their open positions, assessed across every position together.

In practice that means this. Bot A is in drawdown and eating margin. Bot B has a nice winning position open. If your equity drops through the stop out level, your broker closes positions. At Vantage that starts with the position holding the largest loss, and plenty of other brokers work the same way. So bot B's winner is last in the queue, but it is in the queue. And its floating profit was the exact amount propping up the equity bot A was burning.

At Vantage the margin call sits at 50% and the stop out at 20%. Check which entity is on your contract. If you trade through an EU regulated entity, the mandatory close out applies once your cash plus open profit falls below half the minimum required initial margin, so you are closed out earlier than at 20%. That 20% is broker policy and not law, so read your own contract specifications. How those levels work is in margin call versus stop out.

There is a second, less dramatic version that costs just as much. Margin starvation. Bot A has eaten all the free margin, bot B wants to take its best setup of the week, and its order is rejected with return code 10019, "there is not enough money". No crash, no alarm. Just a trade you never had.

Magic numbers partition your trades. They do not partition your margin. MT5 has no risk budget per bot.

Are you diversifying, or placing the same bet three times?

This is where most of the money goes, and it is the least understood part.

Three bots on EURUSD, GBPUSD and AUDUSD feel like three positions. When the dollar moves, they are one position with three names.

Brokers publish correlation tables, and those tell you two things at once. Look at the numbers, then look at what is missing.

Two tables from two brokers on the same pairs, differing by 0.05 to 0.10, and neither says over what period it was measured. Defcofx notes itself that the values are approximate averages for educational purposes and may change depending on market conditions and timeframes. That is an honest disclaimer and it is also exactly the problem: correlation is not a property of a currency pair, it is a snapshot.

MetaQuotes uses hard thresholds in one of its own articles on this, and they work well as a rule of thumb. Below 0.35 is excellent. Between 0.35 and 0.60 is acceptable. From 0.60 up it is dangerous, because your strategies are duplicating each other's exposure.

The academic side is sharper than you would expect. There is a peer reviewed paper titled "Are EUR and GBP different words for the same currency?" which concludes that the GBP already behaves as a true euro. That was on 1990s data and answering a different question, but the thrust is usable: two pairs you treat as different can be statistically almost the same thing.

And then the point that matters most. Correlations are not stable, and they rise exactly when things go wrong. Research on diversification shows that correlations in the left tail, meaning during losses, are much higher than in the right tail. Your diversification works best on the days you do not need it.

If each bot risks 2%, what are you actually risking?

Three bots, 2% risk each per trade, all three in the market at once. Here is the arithmetic.

Read the last column first. Your worst case is 6% at any correlation. Correlation does not change how bad your worst day is. Correlation changes how often your worst day shows up.

At 0.9 your three bots are 1.07 bots. That 6% day arrives whenever the dollar makes one decent move, which is often. At 0 you genuinely have three bots, and three unrelated things have to go wrong at once for the same 6%. Same ceiling, completely different frequency.

Small caveat on the maths: a 2% stop is a fixed loss, not a normally distributed return, so that middle column describes the spread of your aggregate result rather than an exact probability. The direction holds, the precision does not.

MetaQuotes arrives at the same place from another angle in its own research. Five EAs each dutifully respecting an individual 2% daily loss limit produce 6 to 8 percent total account drawdown, because one hits its limit and stops while the other four carry on. Their own recommendation for multi-bot setups: a hard 2% daily cap at account level, and 0.5% per trade.

That is the whole problem in one line. Risk is managed per bot, but the consequences are shared. MT5 has no built-in way to make bots share one risk budget. MetaQuotes has published a working MQL5 pattern for it, so it can be built, you just do not get it for free.

What actually diversifies, and what only looks like it?

What diversifies is the type of strategy.

Trend following and mean reversion are structurally biased to disagree. One buys the breakout, the other sells it. Those two can run on strongly correlated instruments and still produce equity curves that do not move together. That is real diversification.

Two momentum bots with different lookbacks are not diversification. They respond to the same regime, half a day apart.

Our own rule of thumb: measure correlation on your drawdowns rather than your returns. The question you want answered is not whether two strategies move together on average, but whether they lose money together. Those are different questions and only the second one can blow up your account.

Practically, if you do want to run several bots: cut the risk per bot every time you add one. Work with a total budget for the account rather than a percentage per bot. And budget around 600 MB to 1 GB of memory per MT5 terminal once charts are open, so 4 GB runs three or four comfortably.

So a separate account per bot?

That solves stop out contagion. Your correlation stays exactly where it was.

The reason is simple: a long bias in the same market is still a long bias, even split across three accounts. Market risk does not know your accounts are separate. What you do gain is that bot A's drawdown can no longer drag bot B's winner down with it, and that you can finally see per bot who made or lost the money. That last one is a real benefit few people mention: on one account all the statistics run together and you cannot judge a bot fairly.

The costs are real too. An EA only runs on the account you are logged into, so you need several terminals in several folders, and therefore more VPS.

And then the trap almost nobody mentions. If you are on funded accounts, the standard advice to give each bot its own account can itself be a violation.

FTMO allows EAs, with platform limits of 200 orders at a time and a maximum of 2,000 positions per day, and forbids EAs causing more than 2,000 server requests per day. The capital cap is $400,000 per trader or per strategy, and FTMO reserves the right to suspend accounts if identically traded strategies are detected across multiple accounts and the total exceeds that.

FundedNext allows EAs for a fee on MT4 and MT5, and prohibits them on cTrader and Match-Trader. Each EA is required to employ a distinct strategy, there is a maximum allocation of $300,000 per bot strategy, and hedging across multiple accounts is prohibited.

The5ers is strictest: EAs from a provider whose source code you do not own are not allowed, which in practice rules out most commercial EAs.

In other words, the advice you find on every forum can be the exact rule you break at a prop firm. More on that world in propfirm trading in 2026, honestly.

So, multiple bots or not?

Multiple bots can work, but only if you add up risk at account level instead of thinking per bot. Three bots at 2% each add up to one 6%, and at high correlation that day comes around often.

If you do run several: pick different strategy types rather than different pairs, cut the risk per bot with every bot you add, and keep one hard daily cap for the whole account. Same logic as in risk management for a trading bot.

And if one of your bots suddenly stops working after you added a second, that is often not a risk problem but a magic number conflict. We solve that in trading bot not placing trades.

Sources: MetaQuotes wrote about centralised risk management for multiple EAs and about correlation thresholds in a multi-EA portfolio. The FCA rules on CFDs describe the close out at 50% of required margin.