By Lamon Rutten

Algorithmic Trading Overview

Algorithmic trading, also called robot trading, has contributed to losses worth over a trillion dollars in the global financial markets, including the Flash Crash of May 6, 2010 in the US market. However, it has also brought higher liquidity, which reduces bid-ask spreads and enhances the ease at which people can trade, all to the benefit of active traders.

How can one make best use of the merits of algorithmic trading and how should one limit its risks? A balanced approach will requires an industry-wide understanding of the issues faced by market players and policy makers, and a judicious use of regulations and compliance by the exchange and its regulator. 

Institutional investors, brokers, big trading houses and other market players have continued to innovate in order to reduce cost of trade, manage orders, control risks and prevent vulnerabilities. This is also supported by market regulators, governments and exchanges as they seek liquidity and fairness in the market.

For example, in 1970s the Securities Exchange Corporation (SEC) and New York Stock Exchange (NYSE) started sponsoring Supplemental Liquidity Providers (SLP) and Designated Market Makers (DMM) to provide liquidity in the market. 

Such innovation led to the development and growing sophistication of algorithmic trading strategies. The modern-day algorithmic trading strategies were developed through advancement of internet, decimalization and deployment of high-speed computerized trading facilities with pre-programmed algorithms to facilitate quick orders, speed up trade execution and market liquidity and to deal with market vulnerabilities.

ALGORITHMIC TRADING IN A NUTSHELL

Algorithmic trading can be described as the application of a quantitative model which involves an underlying logic and rules based on a computerized system and programming to execute an action. As such, buyers and sellers will be able to place market orders very quickly, reducing their exposure to volatility and leading to highly liquid trading.

Algorithmic analysis is also able to identify possible mispricing and trade on it. For those who want to place big orders, algorithmic tools make it possible to break these up in smaller trades that can be placed without affecting the market. And using an algorithmic trading system considerably reduces the risk of human error.  

Algorithmic trading engines, which give access to mathematical-based trading, now come pre-packaged with many trading softwares. Users can choose from different strategies and tailor the algos for their own preferences. Artificial Intelligence is often embedded to improve the programme’s decision-making. Assuming that algorithmic trading is done legally, the general public will have better access to exchanges and be able to trade with ease.

For example, ICDX expects that clients will make wide use of the algorithmic trading supports available on its mobile front-end software, the software that enables retail clients to access the Exchange’s growing suite of products.

Algorithmic trading has made markets much more efficient, even though it has contributed to Flash Crashes in the past. For example, research shows that High Frequency Trading (HFT), an algorithmic trading strategy with autonomous high-latency arbitrage, has led to decrease in financial transaction costs, volatility, and buy-sell imbalances. In addition, electronic algorithms with automation and high-speed computers have generally reduced latency, facilitated price discoveries and enhanced liquidity in the market. 

An algorithmic trading engine enacts the instructions inputted by its “owner”, reflecting his trading philosophy. From a retail client’s perspective, algorithmic trading can help considerably in the execution of their trading strategy. It will also make it easier to manage risks (for example, the algo can trade, in response to market opportunities or threats, even if its owner is asleep). 

And it makes it possible for retain investors to better follow the example of more experienced traders. Some trading softwares make it possible to engage in “social trading”, where one can copy in real time, for all or a part of one’s investment portfolio, the trading strategy of a proven successful trader (the World Economic Forum has described it as an empowerment tool for small investors, a low-cost sophisticated alternative to the wealth management services that typically are only available to High Net Worth Investors).

THE RISKS OF ALGORITHMIC TRADING

But algorithmic trading does carry a risk. The main problem with algorithm trading is the potential of erroneous programming which can lead to systemic risks. When one algo decides to sell, this can lead other algos to decide, in a fraction of a second, to sell too, leading to a snowball effect with many market participants forced out of their position because of insufficient margins, and a market crash. This has indeed happened. 

For example, in 2017, the price of Ethereum (one of the leading cryptocurrencies) fell from more than USD 300 to only 10 cents in a matter of minutes on one leading exchange; it was triggered by one large sale which caused the price to fall from USD 318 to USD 224, which in turn triggered 800 stop-loss orders (many because of insufficient margins) which led to the market crash. Markets that crash because of algorithmic trading normally revert back to normal in a matter of minutes, but some market players can be seriously hurt during these minutes.

In the major markets, the risk of flash crashes has become a lot smaller since 2010 because there are now so many different types of trading algorithms, with different buy- and sell-triggers (that now also include external events), different trading logics (e.g., focusing on trend following, or arbitrage, or speculation driven by technical signals) and different time horizons. 

Still, they can happen. Just before midnight on 2 January 2019, when traders were just coming back from New Year celebrations and many markets were still only thinly traded, in just 7 minutes the prices of Yen/USD, Australian dollar/USD and Australian dollar/Yen contracts suddenly moved 5-8%, without any apparent fundamental reason.

FACING THE RISKS OF ALGORITHMIC TRADING

Exchanges nowadays have several safeguards in place to prevent flash crashes. They generally test and approve all algorithmic softwares that are allowed to trade on their platforms. They monitor how many trades come in from each computer connected to its system, and will switch off that connection if the order frequency becomes too high (requiring human intervention to bring that computer online again). 

They have risk management systems in place that prevent the build-up of large, insufficiently covered positions. They also commonly have price limits – once prices hit a limit, trade is halted for a number of minutes so that market participants have the time to consider the situation. 

Exchanges have had more difficulty in ensuring that their trading engines continue providing a level playing field to all users, whether large or small. Algorithmic trading engines rely on price signals for trading decisions, and thus, large investors have spent considerable money to try and get the fastest possible connection between themselves and the exchange – so that their orders are received before those of others. This is by placing the server that hosts their trading software as close as possible to the server that hosts the exchange trading platform.

Exchanges have generally treated this as a new business opportunity, offering rental space in their server centers (this is called co-location). But they need to be careful to keep a balance, and not give a trading advantage to those who can afford co-location that makes the market unattractive to those without such financial means. Despite the complaints by high frequency trading firms that make up a large part of exchange volumes, many exchanges have now started using speed bumps to slow down high-speed trading algorithms.

Regulators do not need to overreact to algorithmic trading – it has many benefits, including for retail investors – but they need to make sure that exchanges have efficient systems in place to deal with trading algos, and that they properly enact their controls. Beyond this, regulators should also consider how they can use algorithmic trading engines to their own benefit, in particular to improve market surveillance.