GUIDE

Forex algorithmic trading: a practical guide

Algorithmic trading at retail scale is not high-frequency trading, and confusing the two leads people to buy infrastructure they do not need for an edge they do not have.

Short answer

Algorithmic forex trading means executing a defined rule set automatically rather than discretionarily. At retail scale it runs on MetaTrader on a VPS, holds positions for minutes to days, and competes on discipline rather than speed. It is not high-frequency trading, which requires colocation and capital no retail account has.

What it is, and what it is not

Algorithmic trading means the rules exist before the trade. That is the whole definition, and it is why the approach can be tested at all — a discretionary decision cannot be replayed against history.

It is not high-frequency trading. HFT competes on microseconds, needs servers colocated with the exchange and capital that retail does not have. Retail algorithmic trading competes on discipline, not speed, and any product selling you retail HFT is selling a category error.

The strategy families that actually get used

Family	The idea	Fails when
Trend following	Enter with an established move, exit when it stalls	Markets range — many small losses
Mean reversion	Fade moves that stretch too far from average	A genuine trend begins
Breakout	Enter as price leaves a range	The break is false, which is often
Session or time based	Trade a specific window with known behaviour	The pattern that defined the window decays

Every family has conditions that suit it and conditions that punish it. This is why regime detection matters more than strategy selection, and why our own engine runs several strategies and scores them against recent performance rather than committing to one — the reasoning is on the engine page.

The costs nobody mentions

A machine that stays on. MetaTrader automates only while the terminal runs, so a VPS is a recurring cost.
Data quality. Backtests are only as good as the tick data behind them, and free data is frequently poor enough to invent results.
Spread and slippage. Model them honestly or your tested edge does not survive contact with a broker.
Your own patience. The hardest part of running an algorithm is not switching it off during a normal losing stretch.

Build or buy

Building teaches you more and costs months. Buying is faster and requires you to judge a vendor you cannot inspect. Neither is wrong, and the honest comparison is on building your own EA versus buying one.

If you are buying, the seven-point evidence test is the fastest filter. If you are testing anything at all, do it in the strategy tester and on demo before it sees money.

Common questions

What is algorithmic trading in forex?

Using a program to decide and place trades according to rules defined in advance, rather than deciding each trade yourself. In retail forex this almost always means an expert advisor running inside MetaTrader. The defining feature is that the rules exist before the trade rather than being formed in the moment, which is what makes the approach testable at all.

Is algorithmic trading profitable for retail traders?

It can be, but not for the reason most people assume. The edge is rarely a cleverer strategy — it is consistent execution of a mediocre one, without the hesitation, revenge trading and moved stops that cost discretionary traders far more than strategy choice does. Expect an automated system to make the same strategy perform closer to its tested result, not to invent a better strategy.

Do I need to know how to code?

To build your own, yes — MQL5 for MetaTrader 5, and it is close enough to C++ that the transition is not severe for a programmer. To run somebody else’s, no. Most retail algorithmic traders buy or download an EA and configure it. Understanding what the code does is valuable even then, which is why a vendor who explains their logic is worth more than one who calls it proprietary.