Automated trading is trading in which a computer program decides when to buy or sell and sends the orders itself, following rules set in advance. A person designs and supervises the system but does not approve each trade. The practice depends on electronic trading, since a program can only act on a market that accepts orders electronically.
The term overlaps heavily with algorithmic trading, and the two are often used as synonyms. Where a distinction is drawn, algorithmic trading can include programs that only work an order a human has already decided to place, while automated trading stresses that the whole chain, from signal to order to exit, runs without manual steps. Automated systems now account for a large share of volume in equity, futures and foreign exchange markets.
How it works
An automated system has four linked parts. A data feed supplies prices and other inputs. A strategy component applies the rules of a trading strategy and produces signals. An execution component turns signals into orders and manages them until they are filled or cancelled. A risk component checks every order against limits before release and watches positions and losses afterward.
These parts are usually assembled on an algorithmic trading platform. Before going live, the rules are tested on historical data through backtesting and then run with small size. Once live, the system reacts to each new price or event, so its speed is limited by its software, its hardware and its distance from the exchange.
Main forms
- Execution algorithms. Brokers and investors use programs to split large orders into small pieces over time, often aiming at a benchmark such as the volume-weighted average price. The choice of method is the execution strategy.
- Automated market making. Firms post buy and sell quotes continuously and adjust them as prices move, earning the spread. See market making.
- High-frequency trading. High-frequency trading firms hold positions for very short periods and compete on speed.
- Systematic investing. Funds run rule-based strategies such as trend following or statistical arbitrage over days or months, a field known as quantitative trading.
- Retail automation. Individuals run scripts on broker platforms or use crypto trading bots on digital asset exchanges.
Benefits and costs
A program applies its rules the same way every time. It does not tire, hesitate or act on fear, and it can watch thousands of instruments at once. It can be tested before money is committed, and its decisions leave a full record. For markets as a whole, automation has coincided with narrower bid-ask spreads and lower commissions in the most active instruments.
The costs are less visible. Systems are expensive to build and maintain. A strategy that worked in testing may fail live because the test was flawed or the market changed. Automation removes human error from individual trades but concentrates risk in design and operation: a single faulty setting is repeated on every order until someone intervenes.
Risks and notable incidents
The most cited example is Knight Capital. On 1 August 2012, a faulty software release caused the firm's systems to send a flood of unintended orders to the New York Stock Exchange. In about 45 minutes it built up large unwanted positions and lost roughly $440 million, and it was taken over soon afterward. On 6 May 2010, the flash crash saw US share indices fall sharply and recover within minutes, with automated selling and the withdrawal of automated liquidity both playing a part.
Smaller failures are more common: a fat finger error in a parameter, a runaway loop that repeats an order, or strategies at different firms reacting to each other. Automated systems can also be used for abusive practices such as spoofing, which is illegal in major markets.
Regulation and controls
Regulators do not ban automation but require controls around it. In the United States, the Securities and Exchange Commission adopted the market access rule in 2010, which obliges brokers to apply pre-trade risk checks to all orders sent under their identity. In the European Union, MiFID II, in force since January 2018, requires firms using algorithms to test them, keep order records, and maintain controls including the ability to stop trading at once. Exchanges add their own safeguards, among them price bands, order-rate limits and the circuit breaker.
Inside a firm, the main defense is a risk management system independent of the strategy code, with limits on order size, position, loss and message rate, and a kill switch under human control.