What is algo trading? A plain-English guide

By the AKQX QuantumTrade team · Educational content, not investment advice. Example numbers are illustrative and describe method, not results.

Key takeaways

• An algorithm is just a rule written precisely enough that a computer can follow it.
• Automation removes hesitation and inconsistency. It does not add an edge.
• A flawed rule is followed perfectly, at speed, for as long as it runs. Limits have to be built in.
• You can learn all of it with no real money, using historical data and virtual capital.
• Rules about what is permitted differ by country and change, so check your regulator and broker.

What does an algorithm actually do?

Strip away the jargon and an algorithmic strategy is four decisions made in advance. When to enter. When to exit, both if it works and if it does not. How much to trade. And what to do when something is wrong. A person trading by hand makes those four decisions again for every trade, usually under pressure. An algorithm makes them once, when you write the rule, and then applies them the same way every time.

The glossary entry has the one-sentence definition. What matters in practice is that the rule has to be stated precisely. “Buy when it looks strong” is not a rule. “Buy when the close is above the highest high of the last 20 candles, risk 1% of capital, stop at the lowest low of the last 10” is.

How does a rule become an order?

1. Data arrives. Prices come in from an exchange, as candles of a chosen size or as individual trades.
2. The rule is evaluated. The program checks whether the entry conditions are met, given everything seen so far.
3. A size is calculated. From the risk you allow and the distance to the stop, plus any limits such as a leverage cap.
4. An order is sent, to an exchange or broker, usually through an API key you control.
5. The position is managed. Stops, targets, trailing exits and loss limits act on their own.

In testing, steps 4 and 5 are simulated. In a backtest the data is history and the fills are assumed. In paper trading the data is live and the fills are simulated. Only in live trading do real orders leave the building.

What is it good at, and what is it bad at?

Good at

• Consistency. The same rule gets applied on a Tuesday afternoon and after three losses in a row.
• Testing before risking. Because the rule is explicit, history can tell you whether it was coherent.
• Discipline on exits. Stops and limits fire without anyone having to feel good about it.
• Watching continuously. It does not need sleep or a screen.

Bad at

• Finding an edge. It applies a rule you already believe in. If the belief is wrong, so is the outcome.
• Noticing the world has changed. It will keep taking a setup after the regime that suited it has ended.
• Judging context. Anything that depends on feel has to be reduced to a number, and some of it cannot be.
• Surviving plumbing failures. A stale data feed or a rejected order is a risk of its own.

A worked example, with made-up numbers

Illustrative only. A trader writes a rule: enter long when price closes above the previous 20-candle high, risk 1% of capital, exit at twice the risk or at a stop. Backtested on a year of 1-hour candles it produces 90 trades, 36 winners and 54 losers, with winners averaging 2.0 units and losers 1.0.

The win rate is 40%. Expectancy is 0.40 times 2.0 minus 0.60 times 1.0, which is 0.8 minus 0.6, so 0.2 units per trade before costs. That is positive but thin, and if fees and slippage cost 0.15 units per trade, most of it is gone. The algorithm did not create that thin edge and automation will not widen it. What the test did was tell the trader, with no money at risk, that the idea is marginal and needs a better exit or a better entry before it deserves a paper run.

Five myths worth dropping early

• “You have to be a programmer.” Code is one way to write a rule, and visual builders are another. What you need is the ability to state a rule precisely. The programming is the easy part, and the precision is the hard part.
• “Algorithms win because they are faster.” Speed matters to a small group competing on microseconds. Most individual algorithms act on candles from one minute to one day, where the advantage is consistency, not speed.
• “If the backtest works, the algo works.” A backtest is a simulation of one past with estimated costs. It is the first check, and the weakest, because it is the most optimistic.
• “Once it is running you can forget it.” Data feeds go stale, orders get rejected, margin runs short and markets change character. An unattended algorithm needs alerts and limits, not faith.
• “More indicators make a smarter algo.” Every extra setting is another way for the strategy to fit history's noise. Simple rules tested properly tend to be easier to trust than elaborate ones tuned until they look good.

The sensible way to start

1. Learn the vocabulary. The glossary defines the terms in plain English.
2. Write one simple rule. Two or three settings, stated in a sentence.
3. Backtest it with realistic costs. Follow the backtesting guide.
4. Try to break it. See how to avoid overfitting.
5. Paper trade it. Compare the live behaviour with the backtest. See the three stages.
6. Add the guards. Stops, sizing, and a loss cap and circuit breaker.
7. Only then, and only if you choose, consider real money, in a small amount you can afford to lose entirely.

What this guide does not cover

It does not tell you what is legal or permitted. The rules for algorithmic trading differ between countries, between products and between exchanges and brokers, and they change over time. Before using any automation with real money, check what your regulator, your exchange and your broker currently allow, and speak to a qualified adviser. AKQX is software only and cannot answer that for you.

Questions about what is algo trading

Keep reading