The complete guide to trading bots: what they are, what they cost, what breaks

A trading bot is a program that executes a trading plan without you. That sentence contains everything true about the category, and roughly none of what gets said about it.

What follows is the whole picture in one place: the families of things people call bots, what's inside one, the five numbers that tell you whether it works, what it costs to run, the two ways to fund it, what breaks, and how to tell a real system from a landing page.

What a trading bot is, precisely

Software connected to a trading platform, holding a set of rules: under these conditions, enter; place the stop here and the target there; close when either is hit. The program watches market data, and when the conditions line up, it acts.

The important word is rules, not robot. A bot has no intuition and no opinion. It does what the plan says every time, including the times you'd have hesitated and the times you'd have doubled down. That is simultaneously its entire value and its entire danger, depending on the plan. The long version is in what a trading bot actually is.

The families that exist

"Trading bot" covers at least six unrelated things with unrelated risks. Most of the category's bad reputation comes from people evaluating one and getting burned by another.

None is a scam by definition and none is a guarantee. What separates them is how much you can verify before paying, and how much they demand from you afterward. The last row is where most current marketing lives: a system described as "AI" is usually either a genuine statistical model that nobody can interpret, or ordinary rules with a fashionable label.

What's inside one: data, decision, order, log

Every serious automated system, whatever it trades, is the same four blocks in a row. If a vendor can't describe theirs in these terms, that tells you something.

Note what isn't in that list: prediction. A good system doesn't forecast where the market is going. It finds a repeatable condition where the payoff is skewed in its favor, takes the trade with a defined loss, and repeats.

The five numbers that tell you if it works

Every performance claim reduces to these. Learn them once and most sales material becomes transparent.

Two of those deserve extra emphasis. Win rate is the most abused number in this industry: a system can win 46% of the time and be excellent, or win 85% of the time and be a slow bankruptcy. What matters is win rate combined with the win-to-loss ratio. At a target set to three times the risk, breakeven sits at winning one trade in four, about 25%, and everything above that is edge.

Sample size separates evidence from anecdote. Fifty trades tell you almost nothing; a strong-looking fifty-trade record happens by chance constantly. Hundreds or thousands of trades across multiple years and market regimes is a different class of claim, and the only kind worth paying for.

The arithmetic of a 3-to-1 target

Work it through once and losing streaks stop feeling like failures. Risk one unit to make three. Take a hundred trades at a 40% hit rate, as a working hypothesis and not a promise: forty winners at three units is 120 units, sixty losers at one unit is 60 units. Net 60 units before costs, from a system that was wrong 60% of the time.

Now the consequence people don't want. If you're wrong 60% of the time, runs of four and five losses are ordinary. Not a broken system, not a changed market, just arithmetic. Any equity curve produced this way is a staircase with real valleys in it, and the drawdown figure tells you how deep they've historically been.

The system doesn't have bad weeks. It has the bad weeks its statistics always implied.

What it actually costs to run one

Vendors quote the license price. That's rarely the whole bill.

What breaks

Systems fail in a small number of well-known ways. Knowing them is most of what separates an informed buyer from a hopeful one.

1. Curve fitting. The most common. Parameters tuned until they fit history perfectly, which guarantees a beautiful backtest and no future performance. Symptoms: many parameters, oddly specific values, results that collapse when the dates shift slightly.
2. Regime change. A market's character genuinely shifts: volatility regime, liquidity, participant mix. A system built for one regime degrades in another. This is real and unavoidable; the defense is monitoring, not prevention.
3. Slippage and fills. Backtests assume you got filled. Live markets, especially in the seconds after a data release, do not always agree, and a system whose edge is a tick or two wide can be consumed by that gap.
4. Infrastructure. Power cuts, dropped connections, Windows updates at the wrong moment, an expired contract month. Boring, frequent, and the reason people run on a VPS.
5. The owner. Statistically the biggest one. Switching it off during a drawdown, changing parameters mid-flight, adding contracts after a good month. Covered in automation mistakes and why most traders lose.

How to tell a real one from a toy

• Is there a reproducible backtest? Not screenshots. A strategy you can load into a standard tool like the NinjaTrader 8 Strategy Analyzer, over the stated dates, that returns the stated numbers on your machine.
• Is the drawdown published? A vendor showing profit without showing the worst loss is selling you half a picture, and it's the wrong half.
• Are commissions included? Gross-only figures are a red flag on a high-frequency system and sloppy on any other.
• How long is the record, and how many trades? Multiple years and thousands of trades across different market conditions. Under a few hundred trades is noise.
• Does it say when it doesn't trade? Honest systems stand aside under defined conditions. Systems that always have a position are usually indifferent to them.
• Is the risk defined at entry? Stop and target placed in the market with the entry, not "managed" afterward.
• Are losing months shown? If the marketing has none in it, the marketing is fiction.

Automatic disqualifiers: guaranteed returns, fixed monthly percentages, "risk-free," account screenshots you can't audit, testimonials without verifiable identities, and urgency pressure on the purchase. No legitimate system needs a countdown timer.

Who this isn't for

It does not make sense in four cases, and they're worth being blunt about:

• You want excitement. A good system is deliberately boring. If you're here for the feeling, you'll interfere with it inside a month.
• You need the money. Capital you can't afford to lose changes every decision you make about the system, and all of the changes are bad.
• You expect consistency month to month. The record shows losing months. There will be more.
• You won't leave it alone. If you know you'll tinker, you're buying a track record and then throwing it away. Trade manually instead, and at least the results will be yours.

Frequently asked questions

Are trading bots legal in the United States?

Yes. Automating orders in your own account is legal. What is regulated is managing other people's money or advising professionally for compensation, which requires registration with the CFTC and NFA. Buying software and running it on your own account is neither. Your results still have to be reported to the IRS, and US futures get specific treatment under Section 1256.

Do I need to know how to program?

No. Installing a compiled strategy into NinjaTrader 8 is a matter of importing a file and enabling it on the right account and instrument. Programming knowledge only matters if you intend to build or modify systems yourself, and modifying a purchased one is usually a mistake, since the track record you paid for no longer describes what you're running.

How long before I know whether a system works?

Longer than most people are willing to wait. A statistical edge plays out over dozens or hundreds of trades and can easily start underwater, so a losing first month proves nothing either way. Judge a system by whether it is behaving in line with its own historical statistics, particularly its drawdown profile, not by whether it made money this week.

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.