What Is Algorithmic Trading? A Plain-English Guide

Algorithmic trading means letting a set of precise, written-down rules decide when to buy and sell, instead of deciding trade by trade yourself. The rules can be as simple as "buy when the price crosses above its 50-day average, sell when it crosses back below" or as involved as a portfolio of signals weighed against each other. What makes it algorithmic is not complexity. It is that every decision is defined before the trade happens.

How it differs from manual trading

A manual trader watches charts, forms a view, and clicks the button. That works, but it has two well-known failure modes.

The first is inconsistency. The same setup on a calm Tuesday and after two losing trades on a Friday afternoon rarely gets the same decision. Emotions, fatigue, and recency bias all leak into the judgment.

The second is coverage. A person can genuinely follow a handful of markets during the hours they are awake. Rules running on a server follow every market they are pointed at, around the clock, and react in milliseconds.

An algorithm has neither problem. It applies the same rule the same way every time, and it never sleeps. What it gives up is discretion: it will also apply the rule in conditions you never anticipated, which is why testing matters so much. That trade-off, consistency for judgment, is the core of the whole discipline.

What a trading algorithm is made of

Strip away the vocabulary and nearly every strategy has the same four parts:

1. A universe. Which instruments it trades: one crypto pair, a basket of large-cap stocks, a currency cross.
2. An entry rule. The exact condition that opens a position, usually built from indicators like moving averages, RSI, or price breakouts.
3. An exit rule. The condition that closes it: a profit target, a stop loss, an opposite signal, or simply time.
4. Position sizing. How much to risk on each trade, which does more to determine long-term results than most entry rules do.

If you can state all four without ambiguity, you have a strategy that can be tested and automated. If you cannot, you have a hunch.

The main styles of algorithmic strategy

Most strategies you will encounter fall into a few families:

• Trend following buys strength and sells weakness, betting that moves persist. It loses often and small, and wins rarely and big.
• Mean reversion does the opposite: it bets that stretched prices snap back toward their average. It wins often and small, and must control the rare large loss.
• Breakout strategies wait for price to leave a defined range and ride the expansion that can follow.
• Momentum rotation ranks a basket of instruments and holds the recent leaders, rotating as the ranking changes.

None of these is "best". Each has market conditions where it shines and conditions where it bleeds, which is why understanding backtesting comes immediately after understanding what a strategy is.

What you actually need to start

A decade ago the honest answer was: a programming language, a data vendor, a broker API, and months of plumbing. That barrier has mostly fallen. Today the list is shorter:

• A strategy idea stated precisely. The four parts above.
• Historical data to test it on. You need to see how the rules would have behaved across different market conditions, not just last month.
• A way to execute it. Either code you write and host yourself, or a platform that turns rules into live orders at a connected broker or exchange.

On Horizon you describe the rules in plain English, the platform builds the strategy, backtests it against years of historical data, and can deploy it to trade automatically. If you would rather start from something proven than from a blank page, the strategy templates directory contains ready-made strategies with their historical results attached.

What algorithmic trading is not

A few myths are worth clearing early, because they cause expensive mistakes:

• It is not guaranteed profit. An algorithm executes your rules perfectly; if the rules lose money, it loses money perfectly too.
• It is not only for institutions. Execution speed matters enormously for high-frequency firms, but most retail strategies trade on timeframes of minutes to days, where a well-tested idea matters far more than microseconds.
• It is not set-and-forget. Markets drift. A strategy that worked for two years can decay, so live strategies need monitoring and periodic re-testing.

Where to go next

The single most important skill in this field is judging whether a rule set actually has an edge before risking money on it. That is what backtesting is for, and it is the subject of the next lesson: How to Backtest a Trading Strategy.

Frequently asked questions

Is algorithmic trading profitable?

It can be, but the algorithm is not what makes it profitable. Automation executes whatever rules it is given with perfect discipline; if the rules have no edge, it loses money with the same discipline. Profitability comes from testing ideas honestly and sizing them conservatively, which is what the rest of this curriculum covers.

Do I need to know how to code?

Not anymore. Writing your own engine still requires programming, but platforms now turn rules described in plain English into runnable, backtestable strategies. What you cannot skip is the thinking: stating your entry, exit, sizing, and universe precisely enough to be tested.

How much money do I need to start algorithmic trading?

Learning costs nothing: backtesting and paper trading do not require capital at all. When you do go live, start with an amount whose total loss you could shrug off, and size positions so a single trade risks around 1 to 2 percent of the account.

Is algorithmic trading legal for retail traders?

Yes. Retail algorithmic trading is legal in major markets, and brokers expose APIs precisely so software can trade accounts. What matters is using a regulated broker and respecting its terms; the rules that govern your trading are the same ones that apply when you click the button yourself.