Test your trading bot live on a cent account

The order most people follow: backtest first, then demo, then straight to a thousand on a live account. And that is where it breaks. Because the first time your bot trades real money you discover things that never showed up on demo, and by then it is tuition.

There is a step in between that you can skip but should not.

What exactly is a cent account?

A cent account is a real account where your balance is shown in cents instead of whole euros or dollars. Deposit 20 and your screen reads 2,000. One lot there is 1,000 units, which on a standard account is 0.01 lot. Everything is a hundred times smaller, while the execution is real.

That last part is the whole point. Your orders take the same route, over the same server, through the same spread movements as somebody trading ten thousand. Only your stake is tiny.

In practice that means this: your bot sends exactly the orders it always sends, you see exactly the same behaviour, and a 3 percent loss on the day costs you 60 cents instead of 30. The psychology stays real, because it is your own money. The damage does not.

Why demo does not tell you what you need to know

A demo does not send your orders into the market. The broker fills them internally, neatly, at the price you asked for. No delay, no slippage, no requote.

Live it works differently. Your bot sees a setup, sends an order, and between that moment and the fill there is time. On a quiet EURUSD market you will not notice. At 2:30pm on an NFP Friday, or at the London open on gold, that price can sit a good distance from where your bot asked.

You almost never see that gap on demo, and it is exactly why people run profits on demo and losses live. We dug into the pattern separately in profitable on demo, losing on live. A cent account is the cheapest way to measure that gap before it costs you.

A demo tests your strategy. A cent account tests your execution. Those are two different things and the second one costs people more money.

Setting the test up with twenty

Open a cent account at your broker and deposit an amount that will not keep you awake. Twenty to fifty is fine. Under ten gets awkward, because there is no room for a normal losing streak and your test ends in a margin call instead of a conclusion.

Then set your risk per trade to the same percentage you will use on your real account. If you normally run 1 percent, run 1 percent here. On 20 that is 20 cents per trade. The point is not what you earn, it is that your bot makes the same decisions at the same proportions.

Put it on a VPS after that and leave it alone. Run it from your laptop and you miss every trade that happened while the lid was shut, which tests your sleep schedule rather than your bot. Why that matters so much is in why a VPS matters for your trading bot.

And the thing you really do not do: change something halfway. Every time you touch a setting, your count starts over. That feels unsatisfying, but a test you fiddle with produces no data, only a feeling.

What a cent account does not show you honestly

The costs. Cent accounts often carry a wider spread than a standard account at the same broker, sometimes up to one and a half pips on majors, and the overnight swap still ticks along. Budget roughly 0.10 to 2 per lot per night on anything you hold.

That changes how you read your result. If your bot breaks even on a cent account, that is not bad news, because on a normal account with a tighter spread the same sequence would probably have shown a profit. The reverse is more dangerous: if you post lovely numbers on a cent account with a wide spread, your edge is thick enough to survive costs, and that is precisely what you wanted to know.

So treat the result as a floor and not a forecast. A few other things differ: maximum position size is lower, deposit and withdrawal fees run higher at some brokers than on a standard account, and the support tier you land in is usually a step down. For a test of a few weeks none of that matters.

How long should you let it run?

Until you have at least 60 trades, and preferably closer to 100. Below that you are looking at chance rather than at your bot.

On a bot that trades a few times a day you get there in two to four weeks. On a swing bot that acts twice a week it takes months, and then a cent account mainly tests whether your setup holds: does the connection stay stable, does the VPS keep running, do the orders execute correctly. That is worth knowing too, it is just a different question from whether your strategy makes money.

At the end, do not look at the return in percent. Look at the difference between the price your bot asked for and the price it got, at how many trades were missed, and at whether average win and average loss sit in the same ratio as in your test. Which numbers to use for that is in how to judge a trading bot's results.

When do you move to a standard account?

When you have the trades, the execution threw up no surprises, and the ratios match what you expected. Not when you had a nice week.

The switch itself is boring, and that is how it should be. You open a normal account in your own name, deposit an amount where your risk rules have room to work, and start the same bot on the same settings. How much that needs to be at minimum we worked out in how much money you need for a trading bot.

One thing does change, and it is worth watching: your position size goes up by a factor of a hundred. Trades that filled cleanly on the cent account can pick up a little more slippage on a real account because more volume has to go through. At normal sizes on majors you will barely notice. At thousands on a thin instrument you will.

We run our own members through this order because nothing is attached to it. The account is in your own name, you change your MT5 password after the connection is made, and we take 30 percent of your profit and nothing when you lose. So there is no contract forcing you to scale faster than you want, and no reason for us to push you toward a number you are not ready for.

Twenty and three weeks of patience. That is the entire price of knowing whether your bot does live what it did on paper.