AI Trend Research

← AI glossary

AI glossary · Basic · Classic ML, recsys & search · Reinforcement learning · Finance & insurance

Algorithmic trading

Algorithmic trading is a way of trading in which a computer automatically places orders, leaving decisions about when, what, at what price and how much to buy or sell to predefined rules or models.

About this term
Level
Basic · Technical terms you often meet in AI news
Category
Industry AI
Added
2026-10-10
In plain words Instead of a person watching the market and deciding each time, it is like giving a machine a recipe that says 'when this happens, do this' and having it follow along. Unlike a recipe, though, the market keeps changing, so a rule that worked well yesterday may cause losses today.

How it works

First, a strategy is designed that takes in data such as prices, trading volume and news and produces trading signals. The strategy's performance and risk are checked through backtesting, which applies it to historical data. In live trading, the system automatically places orders when a signal appears, and large orders may be split into small pieces to reduce market impact. Risk management rules such as loss limits and position sizes always operate on top of this.

Components of algorithmic trading

Risk managementLoss limits, position sizes, kill switch

Order executionSplitting large orders into small pieces

Strategy and signalsRules or models, checked by backtesting

DataPrices, volume, news

From the top down, data becomes signals and is executed as orders, while risk management wraps around the whole.

What AI does

Traditional algorithmic trading follows human-made rules, while machine learning is used to find patterns related to price movements in the data on its own. Time series forecasting models may predict short-term prices, and natural language processing may read the sentiment of news and corporate disclosures for use as signals. There is also research using reinforcement learning to optimize how orders are split and placed. However, financial data is noisy, so plausible-looking patterns are often just coincidence.

Things to know

Strategies fitted too closely to historical data easily fall into overfitting and fail in real markets, and data leakage, where future information slips into a backtest, is another common pitfall. When many algorithms react in similar ways, market volatility can suddenly spike, so exchanges and regulators impose rules and safeguards on algorithmic trading. A single system error can lead to large losses in a short time, so kill switches and human oversight are important.

Example

When asset managers buy or sell large blocks of stock, they use algorithms to spread orders over the course of a day so the price does not move suddenly. Individual investors can also use automated trading features in brokerage apps for simple rule-based trades such as 'sell when the price hits this level.'

Related terms

Time series forecasting Reinforcement learning Overfitting Data leakage Robo-advisor Alternative data