The going-live checklist: what to check before you switch a system on

It is 8:29 AM ET. The strategy is enabled, the chart is open, and the connection indicator in the corner is gray instead of green because the platform lost the data feed twenty minutes ago and nobody was watching. The trade the system was supposed to take does not happen. Nothing crashed, nothing errored, and the report at the end of the month will simply be missing a day.

That is what going live badly looks like. Not a dramatic loss, usually. A quiet mismatch between what you tested and what is actually running. Here is the check to run before you switch any automated system on for real money, in the order that matters.

1. The machine, and whether it will still be awake

An automated strategy needs the computer running, the platform open and the connection alive at the moment the window opens. Obvious, until you list the ways it fails: the laptop sleeps, Windows reboots for updates at 8:25 AM, the connection drops for ninety seconds, the platform disconnects overnight and never comes back.

Set the machine to never sleep, move the update window to a time you are not trading, and confirm the platform is connected before the session rather than after. If you are not willing to check that every morning, that is the argument for renting a server, laid out in whether you actually need a VPS.

2. The strategy is running where you think it is

In NinjaTrader 8, a strategy applied to a chart and a strategy enabled from the Strategies tab of the Control Center are not the same thing, and the second is the one that survives a chart being closed. Confirm which you are using, that it points at the account you intend, and that it is actually enabled rather than merely configured.

Then check the start behavior setting, which decides what the strategy does about existing positions and historical orders when it starts. Choosing it carelessly is how people arrive at a live position they never intended. The step-by-step is in how to import and run a NinjaScript strategy.

3. Data: instrument, expiry, session, time zone

Four settings, all boring, all capable of quietly ruining a month.

• Instrument. The micro, not the full-size contract, unless you deliberately chose otherwise. The difference in dollars per point is not subtle.
• Expiry. The front month, the one carrying the volume. Running into an expiring contract is a slow way to get terrible fills.
• Session template. Overnight and regular-hours templates produce different bars from the same data. A strategy tested on one and run on the other is a different strategy.
• Time zone. If the platform is not set to Eastern, a 9:30 AM ET rule fires at the wrong moment, and it will still look correct on your screen.

Then load enough history for whatever the strategy calculates before the first bar of the session. A strategy that needs 200 bars and has 40 behaves differently for the first hour of every day.

5. Read the rules of the account you are about to trade

6. Know how to stop it before you need to

Find the off switch on a calm afternoon, not during a fast market with an open position. Two things to establish and write down.

First, how you disable the strategy so it takes no further trades. Second, how you flatten an open position, and whether disabling closes that position or leaves it sitting there. Those are separate actions, and assuming the first does the second is how a "flat" account ends the day holding a contract. Confirm both on simulation, along with what happens to the working stop and target orders when the strategy stops.

You do not need a plan for the good mornings. You need one for the morning the platform freezes with a position open.

7. Set expectations to the drawdown, not the average

With the target at three times the risk, the system does not have to be right most of the time: break-even sits at a 25% win rate. That geometry is what makes it work, but the experience of it is streaks of losses that are perfectly normal and feel exactly like a broken system.

So decide now, in writing, what would actually make you turn it off, and set that threshold worse than the tested maximum drawdown rather than better. Anything vaguer than a number is a decision you will make at the worst possible moment, which is the failure mode running through the ten mistakes that ruin a working system and our risk page.

Frequently asked questions

How long should I run a system on simulation before going live?

Two weeks of real-time simulation is a reasonable minimum for a system that trades once a day, because it gives you around ten sessions of evidence that your machine, connection and settings all behave at the right hour. That is not enough trades to evaluate the strategy itself, which is what the backtest is for. You are confirming the plumbing, not the edge. If anything at all goes wrong during those two weeks, fix it and restart the clock.

Does disabling a strategy close my open position?

Not necessarily, and this is the single most dangerous assumption in automated trading. Depending on the platform and the settings, disabling can leave the position open and cancel the working orders that were protecting it, which is the worst of both outcomes. Test this on a simulation account with a live position so you know exactly what happens, and keep a separate way to flatten manually.

What if my internet drops during the trade window?

If the stop and target were already placed in the market, they sit with the broker and continue to work whether or not your platform is connected, which is one reason a system should place both at entry rather than managing them locally. What you lose is the ability to take a new trade or to modify anything. Reconnect, check the account for open positions before assuming you are flat, and treat repeated dropouts as a reason to move the setup onto a server.

What is the most common going-live mistake?

Educational content, not investment advice. Trading futures involves substantial risk of loss and is not suitable for every investor. Any reference to Algomyth results is hypothetical, simulated over historical data, and shown in full on the performance page; past performance does not guarantee future results.