Automated trading: what it is, what it costs, what breaks

It is 9:29 AM ET on a Tuesday. Nobody is in the room. There is a cold cup of coffee on a desk in a spare bedroom and the person who owns the account is in the shower.

At 9:30:00 the US cash market opens and the day's volume arrives. Seconds later a program on that computer sends three orders in the same instant: an entry, a stop below it, a target three times as far away. Forty minutes later the position closes and the platform writes a line in a log file. The owner made no decisions this morning. He made them months ago, in writing.

That is automated trading with the marketing stripped off. Not a robot with opinions. A plan that executes itself.

What automated trading is, and what it is not

Automated trading means writing rules a computer can execute, connecting the program to a broker, and letting it send real orders without asking you first. The rules are boring on purpose: when these conditions are true, buy this size, stop here, target there, close when either is touched.

Three things it is not. Each one costs people money every year.

• Not passive income. Nobody pays you for owning software. You pick the capital, size the position, read the account rules and decide whether to keep running it while it is underwater.
• Not hands-off forever. Somebody turns the machine on, keeps the platform updated and reads the logs. Very little work. Not zero.

What it removes is the human at the moment of execution. On the evidence, that is where most retail money dies. Not in the analysis. In the doing.

The chain inside the machine: data, decision, order, log

Every automated system, from a hedge fund's to the one in that spare bedroom, is the same four-link chain.

1. Data in. The platform takes the exchange stream and hands each tick to the program. Delayed or dirty data poisons everything downstream.
2. Decision. The code checks its conditions. Almost always the answer is "not now," and doing nothing is correct. A system that must trade daily was built to look busy.
3. Order out. Orders go to the broker's API. A serious system sends the protective orders in the same breath as the entry, so the worst case is defined before anything happens.
4. Log. Every decision, fill, rejection and error is written down: the link that tells you whether the system still behaves like its own history.

The chain contains no opinion, no hesitation, no revenge. You can see it applied to one specific system on how it works.

From an institutional desk to a spare bedroom

Rule-based trading is not new. Floor traders ran mechanical systems on paper in the 1970s. What was expensive then was not the idea. It was the plumbing: a data feed, a line to the exchange, and programmers.

Three things collapsed those costs. Exchanges went fully electronic, so an order from a laptop reaches the matching engine the way an institution's does. Retail platforms shipped full programming environments. And in 2019 the CME launched micro contracts, cutting trade size by a factor of ten and making a $10,000 account a legitimate place to run a system.

What did not collapse is the hard part. Institutions still have better fills, cheaper commissions and full-time researchers. The gap narrowed on plumbing and stayed exactly as wide on thinking, the subject of algorithmic trading.

The tools got cheap. The discipline never went on sale.

The four things sold under one name

"Automated trading" sells four products with almost nothing in common. Most of the money lost in this field comes from buying one while believing you bought another.

None is fraudulent by definition. What separates them is how much you can inspect before paying: a rule-based system runs in your own platform over dates you choose, and a black box does not. More in best trading bots and best trading systems.

By hand or by system: what actually changes

Automation is not about speed. A system trading one window a day competes with nobody on microseconds. What changes is consistency, and consistency has a dollar value.

That last row is the honest one. Automation moves the failure point from your discipline to your judgment about what to run.

Everything you need to run one

The inventory is short, and every line is checkable before you commit.

None of it requires a technical background. The install and first simulated session are in get started; the index futures specifics are in futures trading bot.

What it costs, line by line

Most pages about automated trading talk about returns and go quiet about costs. Costs are the part you can be certain of, so they come first.

Two routes to the capital, worked out in dollars

There are two places the money can come from, and the choice changes almost everything.

The worked example, dollar by dollar

1. Gross: $26,750 over twelve months.
2. Commissions at ≈$1 per micro take about 5%, leaving roughly $25,400 net.
3. Data and VPS at ≈$27 a month: ≈$320.

Now the part the pitch skips. That works out to about $2,100 net a month, an average across a year that contained losing months as well as good ones. It is one account size over one twelve-month stretch, not a monthly rate anyone is owed, and it is a different account and period from the 44-month report above.

How you measure a system, and the disease that fakes it

Total profit is the least informative number in a track record: the destination, nothing about the road.

Losing more often than it wins reads as a defect until you do the math. With the target at three times the risk, breakeven sits at a 25% hit rate: one win in four draws level, and everything above that is profit. That is what a durable edge looks like: unimpressive per trade, positive in aggregate.

Now the disease. A backtest is the most useful tool in this field and the easiest to poison. Curve fitting is a developer nudging parameters until the historical equity curve looks beautiful; the result is a system tuned to noise that no longer exists. Invisible in the report, obvious in live trading three weeks later. Defenses: a long test period, few parameters, and out-of-sample data the developer never saw. See overfitting and curve fitting.

Everything that can break

• The edge decays. Markets change, and a system that worked for years can stop working without announcing it. You detect it by comparing live behavior to the historical distribution, not by feel.
• Power or internet dies. The program stops watching, but stops and targets already resting at the broker stay alive. This is why real orders beat mental ones.
• The plumbing fails quietly. A platform update breaks the build; a lapsed data feed means no ticks and no trades. You may not notice for days if you never read the logs.
• Slippage in fast markets. Fills a few ticks worse than the backtest assumed. Small each time. Not small over 914 trades.
• You. Switching it off after three red days, widening a target because the market feels lively, doubling size to make back a bad week. Each one turns the thing you tested into something with no track record.

Only one of those is a software problem. The rest are human. What a normal bad stretch looks like is on the risk page.

Warning signs, wrong fits, and real timelines

The warning signs when somebody sells you a system

Account screenshots instead of a reproducible report. Monthly percentage promises, which no fund on earth delivers reliably. A win rate above 80%, which almost always means tiny targets and an enormous hidden stop. No maximum drawdown published. Testimonials with stock photos. Countdown timers. And the clearest tell: you cannot run the thing yourself, before paying, over dates you choose. Checklist in trading bot scams and red flags.

Who this is not for

Anyone trading money they need. Anyone who wants excitement: a good system is deliberately dull, and boredom makes people meddle. Anyone who cannot watch an account fall for six weeks without touching it. Anyone expecting to replace an income in the first quarter.

Frequently asked questions

Is automated trading legal in the United States?

Yes. Running a program that places orders in your own account is legal, and it already accounts for much of the volume on US futures exchanges. Registration with the CFTC and NFA applies to people managing other people's money, not to a self-directed trader using software. Your results are still reportable to the IRS, and futures fall under Section 1256.

Does my computer have to be on all day?

Only during the window the system trades. A strategy that works one session a day needs the platform running through that session and nothing more. A small virtual private server keeps everything up around the clock for a few dollars a month.

How do I know whether the system I bought is still working?

You compare it against its own history rather than against your hopes. Four or five losses in a row are normal for a system whose target is three times its risk, and prove nothing. What matters is whether the profit factor and the drawdown stay inside the range the backtest showed.

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.