What is algorithmic trading?

A working definition

Algorithmic trading is the use of computer programs to make and carry out trading decisions according to rules defined in advance. Depending on the system, the program decides some or all of the following: whether to trade at all, which instrument to trade, when to send an order, at what price, in what size, and how to manage that order once it has reached the market.

Regulators describe it in much the same way. The European Union’s MiFID II framework treats trading as algorithmic when a computer algorithm automatically determines individual order parameters, such as timing, price or quantity, with limited or no human intervention. In the United States, the SEC’s 2020 staff report on the subject uses a similarly broad lens and looks at algorithms both as a way of deciding what to trade and as a way of executing trades.

The important point is that the rules exist before the trade. A human still designs the logic, sets the limits and decides when a system runs. What changes is that each individual decision is made consistently by software, at machine speed, and is recorded as it happens.

Two different jobs: deciding and executing

People often use the term for two quite different activities, and it helps to separate them.

Execution algorithms

The decision to buy or sell has already been made, often by a person or a portfolio model. The algorithm’s task is to complete that order at the lowest total cost, for example by splitting a large order into many small pieces over time. VWAP, TWAP and percentage-of-volume algorithms are common examples.

Decision algorithms

Sometimes called signal or alpha models. These analyse market data and decide what position to hold. A trend-following model that buys when prices have risen persistently is a decision algorithm.

Most professional trading operations use both. A decision model produces a target position, and an execution layer works out how to reach it without paying too much in fees, spread and market impact.

Why it became the norm

Automation followed the move from trading floors to electronic order books. Once prices and order entry were available as data over a network, any repeated decision could in principle be written down as code. Tick sizes shrank, trading spread across more venues and the volume of market data grew far beyond what a person can watch. Software became the practical way to cope.

The Bank for International Settlements has documented the same shift in foreign exchange, where execution algorithms moved from a niche product to a routine way for institutions to trade. Digital-asset markets were electronic from the start, and nearly every major exchange publishes an application programming interface (API) for data and order entry.

Who uses it

• Asset managers and pension funds, mainly to execute large orders quietly and cheaply.
• Banks and brokers, which offer execution algorithms to clients and use them for their own hedging.
• Market makers, who continuously quote buy and sell prices and must update them faster than any person could.
• Proprietary trading firms, which trade their own capital with systematic strategies.
• Hedge funds running quantitative strategies across equities, futures, currencies and digital assets.

Algorithmic trading and high-frequency trading

High-frequency trading (HFT) is a subset of algorithmic trading. It is defined by very short holding periods, very high message rates and heavy investment in low latency, such as placing servers in the same data centre as the exchange. Many algorithmic strategies have nothing to do with HFT. A system that rebalances a portfolio a few times a day, or works a single large order over several hours, is algorithmic without being high frequency.

What it is not

Algorithmic trading does not guarantee profit and it is not a machine that predicts prices. A rule that worked in the past can stop working when market conditions change, and a small software fault can cause large losses very quickly. The discipline lies in the process around the code: careful research, honest testing, conservative limits and constant monitoring. The rest of this series covers each of those in turn.