Bots, trade copiers and several accounts at once are all allowed on a futures prop account, on an evaluation and after funding. Three things are not: order rates only a machine could sustain, opposite positions across accounts, and profit that comes mostly from trades held five seconds or less. Every account keeps its own limits, and you answer for what your automation sends.
Evaluation and funded accounts are simulated until the Live stage, and futures trading carries a substantial risk of loss.
What is a bot allowed to do?
Everything a discretionary trader can do, with no declaration and no approval step. Automation is allowed on every account type, from day one of an evaluation through a funded account: the rules engine looks at orders, not at their author. Which also means a bot gets no allowances. Five constraints bind it:
- Flat time: no open position at 4:15 pm ET on Rithmic or 4:45 pm ET on Tradovate, and no entries after it.
- Position limit: 6 minis on a 100K, micros at one tenth, counted across all products at once.
- Daily Loss Limit: $1,800 on a 100K Cycle. It rejects orders until 6:00 pm ET, and a bot with no kill switch keeps firing into the rejections.
- News window on funded Stream accounts: flat one minute either side of a red-folder release, resting orders included. See the release-day article.
- Consistency: a bot that sizes up on a trend day can produce a day large enough to delay a payout on Cycle and Direct.
Rithmic offers a direct API and the usual front-ends; Tradovate runs strategies from its own platform, TradingView or NinjaTrader, within 5,000 actions per 60 minutes. The account rules are identical on both.
Can I copy trades between my own accounts?
Yes. Trade copiers are allowed, mirroring one of your accounts into the others or placing trades from an external signal. One condition: every account involved trades the same direction at the same time.
That condition is the whole rule. Long ES on one account and short ES on another is hedging: a permanent ban on every account involved, copied or manual. It covers correlated products, so long NQ against short MNQ is the same breach in different clothing, and it covers people, so a long on your account against a short on a partner’s is treated identically. The reason is structural: two simulated accounts on opposite sides make a free option, one guaranteed to reach a payout while the other fails for the price of an evaluation.
Copying to accounts that are not yours is where good-faith traders get caught. Selling a signal is not the problem; not seeing a subscriber’s other accounts is. If two of them run your signal inverted between themselves, the positions are the evidence.
How many accounts can I hold?
As many as you buy. There is no published cap, and running a 50K and a 150K in parallel is a common way to size up. Each account keeps its own Max Loss Limit, Daily Loss Limit, position limit and consistency check, and none of them net against each other.
A copier mirroring a master position into three accounts must respect the smallest position limit in the set. On a 25K the ceiling is 2 minis; on a 150K it is 10. Copy a 6-mini master trade into a 25K and part of the order is rejected, leaving a size you did not choose.
The correlation rule is easy to breach by omission: at any moment, all your accounts must be on the same side of a given product. Two bots with different logic will eventually disagree, and the minute one goes long ES while the other is short MES, the flag fires. Parallel strategies work only on different products, or under a single risk layer enforcing one direction per product.
Which practices are prohibited?
| Practice | Status | Condition or trigger |
|---|---|---|
| Automated strategies | Allowed | — |
| Scalping, seconds to minutes | Allowed | — |
| Copiers across your own accounts | Allowed | Same direction everywhere |
| Copiers on external signals | Under conditions | You answer for every order |
| Several accounts in parallel | Under conditions | One direction per product |
| News trading | Under conditions | Funded Stream: flat one minute either side |
| High-frequency order rates | Prohibited | Warning, then closure |
| Micro-scalping | Prohibited | Over 50% of profit from trades ≤ 5 seconds |
| Hedging, one or several accounts | Prohibited | Permanent ban |
Three of those prohibited practices catch traders who were not trying to game anything.
Latency arbitrage. A strategy that reads a faster feed and hits the simulated book before it reprints trades the gap between two data sources, not a market. Few build it deliberately; they buy a “scalping EA” built on that gap, which is exactly why it looks profitable.
Micro-scalping. Trades held five seconds or less. A handful is normal, and nobody counts your fastest exits. The threshold is the share of profit, not the average hold time: over half the account’s profit from those trades. A bot taking two ticks in three seconds hundreds of times a week lands there.
An erroneous price. A simulated feed occasionally prints a quote the exchange never traded. Filling against it and banking the result is a trade the review will unwind. Treat an impossible print as a data fault and tell support, rather than sizing into it.
What actually happens if you cross the line?
The rules engine flags patterns automatically: opposite positions across accounts, hold-time distributions, order rates. A human reviews every flag before a decision, and you are told what was flagged and why.
Then the outcomes split. Hedging is a permanent ban across every account involved, with no warning stage, because the pattern is unambiguous. High-frequency rates and micro-scalping get a warning first and closure only if the behaviour continues: both can be a badly tuned strategy rather than an exploit. Any pending payout is held during the review.
A checklist before you connect an automated strategy
Run these on a fresh account before one with room at stake:
| Check | What to verify |
|---|---|
| Time filter | No entry after 3:45 pm ET; flat before your platform’s flat time |
| Position cap | Hard-coded at the smallest account’s ceiling, all products together |
| Daily stop | Below $1,800 on a 100K, bot disabled for the day when it hits |
| Kill switch | The strategy stops after N rejected orders instead of retrying |
| Order rate | Log orders per minute for a week; no human rate is the line |
| Hold times | The share of profit from trades ≤ 5 seconds, not the average hold time |
| Direction lock | One layer above all accounts enforcing one side per product |
| Copy sizing | Ratios per account, checked against the smallest position limit |
| News calendar | Red-folder filter if a funded Stream account is in the set |
| Bad prints | Impossible quotes rejected, not traded |
Most of these are ten lines of code, and they are the ten lines missing from every bought strategy. The room they protect is finite: a 100K Cycle carries a $3,000 Max Loss Limit against a $6,000 target, so a bot without a daily stop can spend a third of the account in one unattended afternoon. The article on evaluation failures shows that sequence with a human at the keyboard; automation only runs it faster.
Weighing an automated approach against your own capital? The comparison of an evaluation and a personal account sets out what each costs. Sizes, limits and prices are on the pricing page.
Sources
- CME Group, Globex trading hours and product calendars: session times referenced for flat-time planning.
- CME Group, E-mini and Micro E-mini equity index contract specifications: contract sizes used in the position-limit examples.
- Commodity Futures Trading Commission, customer advisories on automated trading systems and signal services.
- National Futures Association, rules on trading standards and account handling.
Simulated accounts until Live. Futures trading involves substantial risk of loss.