Computer algorithms automate financial market asset trading in algorithmic trading. Market data, including price, volume and timing, guides these algorithms to execute transactions at appropriate speeds and volumes. The objective is to make trade choices without human intervention so as to speed up execution, minimise expenses and evaluate massive volumes of data for better strategies. Algorithmic trading accounts for 60-75% of activity in financial markets, which allows high-frequency trading, reduces human error and allows trading methods that were previously inconceivable at such speed and scale.

However, trading algorithms did not appear suddenly - it was conceived in the 1970s by the New York Stock Exchange (NYSE) when computers and technology changed financial markets. Before algorithmic trading, brokers personally called or placed orders for most market transactions. Trading was sluggish and depended on human decision-making so resulting in emotional biases, blunders and delays. However, the real adoption began when the NYSE and other exchanges implemented electronic trading systems in the late 20th century whereby enabling faster trade execution; the rise of computers and the internet in the 1980s and 1990s further accelerated the process. It was also during this period that the first algorithmic trading strategies were developed - these early algorithms were basic, designed to follow a few key instructions such as executing orders when certain price points were reached. As trading volumes increased and markets became more complex, so did the algorithms. By the early 2000s, more sophisticated high-frequency trading (HFT) strategies emerged, allowing firms to place thousands of orders per second. This was made possible by advancements in network speed, data processing power and algorithm design. Financial institutions began to embrace the potential of algorithms to handle the massive amount of data being generated, leading to increased market efficiency and liquidity. Due to its speed, accuracy and bias-free trading, algorithmic trading became popular. Therefore, hedge funds, asset managers and investment banks started optimising their trading tactics with algorithms and this, in turn, heralded the start of a new financial market age when robots could execute transactions and analyse massive data sets in real time.

Algorithmic trading may seem complex but, at its core, it is about automating decision-making through code. The process can be broken down into several key steps which ensure that trades are executed quickly, efficiently and with minimal human involvement:

· defining the strategy - every algorithm begins with a strategy, which governs trading. This can include simple strategies such as buying a stock at a certain price, or more complex ones such as statistical arbitrage which uses market data patterns. Based on real-time market conditions, algorithms should follow these criteria exactly.

· strategy back testing - following strategy definition, historical data is used to evaluate its efficacy. Back testing mimics algorithm performance in the past and, if the method works, it may be used in actual trading. If results are poor, it may need modifications before going live.

· code - after choosing a strategy, the algorithm is coded where skilled programmers write algorithm decision-making instructions; Python, C++ and Java are used to program these algorithms because they are quick and precise.

· market data analysis - the algorithm requires market data, prices, volumes, news, economic statistics, and more. The strategy-based algorithm analyses this information and draws conclusions and the data stream enables software to react fast to changes in the market (the algorithm acts after locating a transaction). People are slower than market orders and so high-frequency trading, which occurs in thousands of transactions per second, uses this approach to maximise gains from small market fluctuations.

In addition, AI is transforming algorithmic trading by making algorithms smarter, faster and more adaptable. One of the most significant advancements is machine learning (ML) which allows algorithms to learn from data and adapt to changing market conditions. This enables algorithms to improve over time, making more accurate predictions and executing trades more efficiently. Predictive analytics powered by AI can analyse large datasets to forecast market movements, giving traders a competitive edge. Moreover, natural language processing (NLP) allows algorithms to process news and social media in real-time, understanding market sentiment and reacting to breaking news or changes in investor behaviour. AI also brings unprecedented speed and efficiency. By processing vast amounts of data in real-time, AI-powered algorithms can execute trades at speeds far beyond human capabilities, which is crucial in high-frequency trading. In terms of risk management, AI systems continuously monitor market data to assess and adjust strategies in real-time, so reducing risk in volatile conditions. AI is also enabling the development of autonomous trading systems that can make decisions and execute trades without human intervention whereby maximising efficiency and minimising errors.

Advantages of algorithmic trading

· speed - one of the biggest advantages of algorithmic trading is its speed. Algorithms can execute trades at a fraction of a second, much faster than any human trader. This ability is particularly important in high-frequency trading where every millisecond can make a difference in profits.

· accuracy - another key benefit is accuracy; algorithms follow predefined strategies without the potential for human error, ensuring that trades are executed according to plan. This eliminates mistakes caused by emotions or manual miscalculations, leading to more consistent results.

· cost efficiency - cost efficiency is also a major advantage. Since algorithms automate the trading process, it reduces the need for human brokers and other intermediaries. This lowers transaction costs and increases overall market efficiency, benefiting both institutional traders and retail investors.

· data-driven insights - algorithms can analyse massive datasets in real-time, providing insights into market trends that may not be visible to human traders. This data-driven approach enables better decision-making and allows traders to capitalise on patterns and opportunities more effectively.

Disadvantages of algorithmic trading

· market volatility - however, despite the many benefits, there are significant disadvantages to algorithmic trading. One of the primary risks is market volatility. Algorithms can sometimes react too quickly to sudden market movements, exacerbating market crashes or flash crashes. In these cases, a large number of trades can be triggered within seconds, leading to massive price fluctuations.

· over-reliance on technology - another concern is over-reliance on technology. Algorithms are only as good as the data they are based on and the strategies they are programmed with. If a market condition arises that the algorithm is not prepared for, it can result in poor decision-making or even financial losses.

· lack of human judgment - lack of human judgment is another limitation. Whilst algorithms are efficient, they lack the intuitive decision-making that human traders bring to the table. This means they cannot account for sudden changes in market sentiment or unpredictable events that might affect the market.

· regulatory challenges - finally, regulatory challenges surround algorithmic trading. As these systems have become more sophisticated, regulators have struggled to keep up. This has led to concerns about market manipulation and the need for stricter oversight to ensure fairness and transparency.

Source: Robinhood

An example of how AI-driven trading is impacting financial markets is the recent announcement that the US broker, Robinhood, will be using AI to offer portfolio management at as low as a fraction of the traditional cost capping fees (in some cases to as low as $250 p.a.) Algorithmic trading boosts global liquidity, price discovery, market effect mitigation and competitiveness; trade execution algorithms boost buying and selling, keeping markets busy and counterparties safe, so the economy benefits from liquidity. Due to real-time data updates, computers adjust prices more precisely where values become obvious and fair. Splitting orders into several transactions helps large institutional investors avoid price disruption and execution expenses; notably, algorithms that compress bid-ask spreads, establish trading strategies and reduce expenses have enhanced competition. This more dynamic and cost-effective trading environment certainly helps individual and institutional investors; however, algorithmic trading is dangerous. Automation might create flash crashes or price volatility during market stress and these events damage investor trust and financial stability, stressing market monitoring and protection. Certainly, caution is required - as Jonathan Hall, an advisor to the Bank of England, pointed out in his speech, ‘Monsters of the deep’ - because technology can create unintended consequences and so undermine confidence in markets. Indeed, recently in the US, a “Fintech founder [was] charged with fraud after ‘AI’ shopping app found to be powered by humans in the Philippines.”

Essentially, algorithmic trading has redefined the markets - not by human intuition, but by pure code. As we see more and more assets such as real estate, bonds, equities and funds being tokenised, it will enable AI powered bots to realise Larry Finks’ (CEO of the world’s biggest asset management company, BlackRock) prediction of mass customisation. But, in order for the real benefits of tokenisation to happen, as CNBC has reported: “One big cybersecurity challenge stands in the way: digital verification.” What once took minutes now happens in milliseconds - executed by machines that never sleep and powered by AI that learns as it trades. It is efficient, precise and largely invisible - until something breaks. These algorithms are not simply tools, they are becoming the market. They move billions, shape prices and decide winners faster than any human can react. And, with platforms such as Robinhood offering AI-powered portfolio management, what was once elite is now increasingly accessible. But speed cuts both ways. Flash crashes, runaway trades and systemic shocks expose a deeper truth: in a world where machines make decisions, oversight cannot be optional - it must be just as intelligent, just as fast. So, the question is not whether AI will dominate trading it already has. The real question is whether the rest of the system can keep up - or if the code will outrun the economy itself.

This article first appeared in Digital Bytes (22nd of April, 2025), a weekly newsletter by Jonny Fry of Team Blockchain.