Automated Trading on FTMO: The Pre-Flight Checklist

Almost everyone searching for automated trading FTMO has already made the decision. The strategy is chosen, the software is bought or nearly bought, and what is left is a quiet fear of losing a challenge fee to a rule nobody explained. This is a checklist for that moment, in the order a trader actually executes it, from the first support ticket to the first live day.

Every item is a question you answer yes or no, and every one of them can be answered tonight. A no does not mean go live carefully, it means do not go live yet. The failure mode in this category is rarely a bad strategy: it is a reasonable strategy running inside a process that skipped a step. If you are still choosing the software itself, our overview of the [best AI trading bots for prop firms in 2026](/blog/best-ai-trading-bot-prop-firms-2026) covers that question. This piece assumes it is settled.

Before you buy the challenge, before you install anything, open a support ticket and ask whether automation is permitted on the exact product you intend to purchase, on the platform you intend to use. Not the brand in general.

Automation is commonly permitted under written conditions across much of the retail forex prop firm market, but conditions differ by account type, by platform and by phase, and they get restructured over time. A permission that applies during the evaluation is not automatically the one that applies once you are funded.

Ask short questions that can only be answered yes or no, then save the dated reply next to the receipt for the challenge. The full method is in [can you use an EA on FTMO](/blog/can-you-use-an-ea-on-ftmo-2026). If the reply is ambiguous, ask again rather than hope.

A robot obeys what it was configured with, not what the rulebook says. So before configuration, put the account's constraints on one page, in numbers, taken from the firm's own current documentation rather than from a vendor summary or from this article. Rules are commonly structured in similar ways across the sector, but the figures differ by account type and change over time.

Four constraints do most of the damage when left implicit. The daily loss limit, because it resets on the firm's clock, not your local time. The maximum loss limit, because it behaves differently depending on whether it is static or trails your equity high, a distinction covered in [daily loss limits versus maximum drawdown](/blog/daily-loss-vs-max-drawdown). The minimum trading days, because a robot that hits the target in four days on an account requiring more has not finished. And any restriction around news releases or weekend exposure, because the software will hold through both unless told otherwise.

This is where most automated evaluations are quietly decided, and it is the step most traders get backwards. The instinct is to work from the target: the challenge asks for a certain percentage, so risk enough per trade to get there quickly. That arithmetic optimizes for speed and ignores the only event that ends the account.

Work the other way. Start from the daily loss limit, decide how many consecutive losses the account must survive inside one day without touching it, and let that division set the risk per trade. As a worked example on a hypothetical account with a five percent daily limit, requiring six losing trades in a day to still leave the account alive puts risk per trade at roughly eight tenths of one percent. Set the software's own stop below the firm's limit, not on it. Your figures will differ, so redo the arithmetic with the numbers from item two.

None of this makes the target unreachable. The target is fixed, so anything above it is surplus, and minimum trading day requirements mean you were never finishing in three days anyway. Taking eleven days instead of six costs you nothing. Touching the daily limit on day two costs you the fee. The longer treatment is in [position sizing for a prop firm challenge](/blog/position-sizing-prop-firm-challenge).

A backtest is a research artifact, not a result. A demo week does not close the gap between the two, but it catches the failures a backtest structurally cannot show: a symbol suffix that does not match, a spread that widens at your entry hour, a server time two hours from the one your settings assume, a minimum lot size that rounds your risk calculation into something else.

Run it on the same broker and account type, with the settings you intend to use, for a full week, then compare the demo behaviour against the backtest over the same period, looking for divergence in trade count and average result rather than in profit. One week is long enough to expose configuration errors and far too short to say anything about the strategy, so treat a clean week as permission to proceed, never as evidence of an edge.

A robot that is not running is not managing anything, and the positions it opened do not close themselves politely when your laptop sleeps. A machine that reboots for an update at three in the morning, a connection that drops for twenty minutes, a lid closed on the way out of the house: each has ended accounts. The usual answer is a server that stays on, close to the broker for execution, with automatic restart and alerting when the terminal disconnects. The options are compared in [choosing a VPS for a prop firm EA](/blog/vps-for-prop-firm-ea).

Then test the failure, because untested recovery is not recovery. Reboot the machine with a position open. If the software resumes and recognises that position, you have an answer. If it opens a second one, you found the problem on demo instead of on a funded account.

Nobody writes this one down, because it feels like planning to fail. Decide now, while nothing is at stake, the condition under which you turn the software off. Not its internal stop loss, which is item three. Your rule, about the whole deployment.

It takes the form of a number and a behaviour. A cumulative drawdown at which you stop and review rather than continue and hope. A pattern, such as the software trading outside the hours it was configured for, that means something is wrong regardless of whether those trades made money. Write both down, with the number, beside the four rules from item two.

Decide it in advance because the alternative is deciding during a losing week, when the honest options are all uncomfortable and the tempting one is to raise the risk and make it back. That decision, taken under pressure, converts more survivable drawdowns into failed evaluations than any strategy flaw.

Disclosure, because what follows is about our own product: PraxAI publishes this blog and sells rule respecting automation for MetaTrader 4 and MetaTrader 5, with a cTrader cBot that is new and still in validation with the first customers, and NinjaTrader 8 for futures, where automation depends entirely on each firm's own policy and has to be confirmed with that firm. Read it as a vendor applying its own checklist.

Three items here can be solved in code. PraxAI GUARD is the drawdown lock: it shuts the robot down before your configured daily limit is reached, so item two stops depending on your attention at two in the afternoon. PraxAI SIZER handles item three, sizing each position from the risk you defined rather than from a lot number carried over from another account. The configuration we validate on gold opens one position at a time, with no martingale, no grid, and a fixed stop loss.

On evidence, we publish 24 approved challenge accounts at praxai.io/results, each with its equity curve and the real maximum drawdown displayed beside the gain, across firms including FTMO, The5ers, E8 Markets and FundedNext. The FTMO 100,000 account there reached 10.28 percent in 12 days with a 1.96 percent maximum drawdown over 42 trades, and the highest drawdown across all 24 accounts is 3.35 percent. Those are results we publish ourselves and you can check the curves on the page. They are not an independent third party audit, and no software makes a pass certain.

The rest of the list stays yours. Nobody can get the written policy answer for your account except you, and nobody else keeps the machine running or respects the stop rule. That is worth remembering when a vendor describes an autonomous system that handles everything, a claim we take apart in [the truth about AI agents on funded accounts](/blog/ai-agents-funded-accounts-truth). Once the account is passed the game changes shape, so [how to keep a funded account](/blog/how-to-keep-funded-account) is the next read: only 1 to 3 percent of funded traders keep the account long term.