Running an automated strategy in Demo mode is easy, and it usually feels good. Moving it to Live mode is a separate decision, and it is rarely treated as separate enough.
The difference in one sentence: in Demo mode you are measuring the logic of the strategy; in Live mode you are measuring the logic plus market friction plus your own behaviour. This post is about testing those three things before you make the switch.
What Demo mode actually measures
Demo mode answers one question well: does the strategy behave the way it was described? Does it open a position under the expected condition, stay out under the unexpected one, and is there a flaw in the logic?
Those are important questions, and Demo mode is the right tool for them. But there are questions it cannot answer:
- Would that order actually have filled? Demo assumes a trade at a price; in a real book there may be no counterparty at that price.
- How much room do fees take? In a high-frequency strategy, trading fees can consume a meaningful share of the return that exists on paper.
- What would you have done? Watching a Demo bot lose for three days running is easy. The same three days with real money is a completely different experience.
The third item is the most underrated. The most common failure mode for automated strategies is not the strategy breaking. It is the user stopping it by hand at its worst moment.
Five questions before the switch
1. Which market condition was this bot built for? Every strategy carries an assumption: a trend, a range, a rise in volatility. If you can name the assumption, you can also recognise the period in which it no longer holds. If you cannot name it, what you have is not a strategy but a set of parameters.
2. How much does it risk per trade? You should know the answer as a percentage. “The whole balance” means one bad trade can end the account. Size a bot by how much it can lose, not by how much it can make.
3. How many trades have you seen? A ten-trade Demo history is an impression, not a result. The share of luck falls as the number of trades rises. For a strategy that trades rarely, that means the Demo period has to be long.
4. What was the worst run? Look at the longest sequence of consecutive losses, not the total. You will live through that sequence in Live mode, and the decision you make during it will matter more than the bot’s performance.
5. What will make you stop it? This is the one sentence that has to be written down before you go live. “I’ll stop it if it goes badly” is not a rule, because the definition of “badly” changes precisely when things go badly.
What starting small actually solves
The recommended way to go live is to run the strategy exactly as it is, but with a small amount. The point is not profit; a small size will not produce a meaningful one.
The point is to measure the three things Demo mode could not: are orders filling as expected, do fees match your assumption, and how do you react to a real loss on screen? The first two become clear within a handful of trades. The third becomes clear in the first losing run.
Skipping this stage and starting at your target size buys you time and nothing else; the cost is that the first surprise is an expensive one.
Write the stop rule in advance
Before switching to Live mode, answer these three and save the answers somewhere:
- What percentage of the balance can it lose before I stop it?
- After how many consecutive losing trades do I stop and review?
- Which observable signal tells me the bot’s assumption no longer holds, and will I stop it when I see that signal?
The value of these three answers is not in finding the perfect numbers. It is in choosing the numbers while calm. If you are forced to make the same decision during a drawdown, what you produce is a reaction, not an analysis.
Control always stays with you: a bot running in Live mode can be stopped whenever you want. That means the rule is enforceable. It does not mean it will be enforced. That part is on you.
One last note
Automation does not remove decisions; it moves them earlier. The decision you make when you take a bot live is the sum of all the decisions that bot will make for months.
So treat the switch as a threshold to pause at, not a step to click through. Demo mode exists for exactly this, and when it is not rushed it is the cheapest learning tool available.
