Gone are the days of a room full of traders frantically executing trades trying to follow volatile market. Computer algorithms are the technology that shape the market today. Up to 70% of all trades in the United States are now performed by machines and not humans.
While algorithmic trading is continuing to grow, the technology keeps improving as well. There are a number of significant technological advancements that are already being implemented and may become a part of the near future in trading. These range from simply improving on precision and speed of trade execution to using complex AI systems to perform market analysis and make accurate trend prediction. In this article we will take a look at the most promising technology to improve and develop algorithmic trading.
Technological progression
Speed
For those unfamiliar with algorithmic trading, it can be defined as any trading that takes place on an automated level. Computers are given specific algorithms for making trades at large volumes and high speeds. The most significant advantage of algo trading is speedy execution. Computers have the ability to execute trades millions of times faster than a human being would be able to execute them manually. It is hard for a person to even comprehend the value of a nanosecond while it is crucial to get ahead of the competing traders.
Faster speed means efficient surveillance of the state of the market and better monitoring of changes in price and volume. Algorithmic trading can potentially help traders execute orders faster, expand strategy portfolios by using more advanced quantitative tools and remove human emotions that often affect the performance of trading strategies.
The future systems that are being developed and tested now will eliminate errors by analyzing the market, historical data and trading history much faster than is currently possible. They will have an ability to employ different trading strategies by spreading the risk across multiple accounts and quickly adjusting to changing market conditions.
The hardware can also be improved, implementing programmable chips for the most efficient and advanced performance where the algorithm will be able to perform multiple tasks at once utilizing a single engine.
Microwave technology
According to Information Week Magazine, “a one millisecond advantage in trading applications can be worth $100 million a year to a major brokerage firm” and that is why the technology keeps on developing and improving its results. Even a small advantage could result in a significantly higher profit rate.
One of the first cases of major speed improvements in algo trading was demonstrated when a high-speed fiber optic cable was built between Chicago and New York back in 2010. As a result of this technology, the communication speed has decreased from 16 milliseconds to 13 milliseconds. Currently, the optic cable technology is being replaced by microwaves. Because the signal travels faster through air than through glass, microwave technology, which uses microwave towers built between major stock exchanges, offers a distinct advantage.
Microwave systems reduce the time required to access the market between New York and Chicago by approximately 40%. The fiber cables require 8.3 milliseconds while microwave ranges from 4.6 to 4.74. As more and more advanced technology is developed and utilized, the speed of trade execution keeps increasing and might soon be reduced to a matter of nanoseconds. According to reports, some HFT firms are able to send order executions in 740 nanoseconds (0.00074 milliseconds).
FPGA (Field-programmable gate array)
Another important factor for speedy trading is implementing advanced hardware which is often delegated to the computer-intensive portions of trading functions such as GPUs, FPGAs, or custom processors.
CPUs are general-purpose and are also used to process a small set of instructions; however, they are not sufficient enough to maintain the required speed of trade execution. FPGAs on the other hand can be reprogrammed to desired requirements after manufacturing, so it’s a flexible alternative to ASICs that have fixed functionality.
Colocation
For high-frequency traders, low latency is important in order to achieve the most profit.The best way to achieve that is to locate servers where algorithms are run as close as possible to the trading platform’s data engines. This process is called colocation – placing servers in close physical proximity, sometimes even in the same room as the exchange servers. This allows trades to be executed without being hindered by the time of space travelling. Also, such colocated servers get access to the quotes and other data (order book, transactions, price, etc.) faster than other market participants. Since speed is a critical factor here, they are able to place orders at the top of the queue at any price. Colocation has become a profitable business for exchanges which charge millions of dollars for the low latency access.
The London-based exchange BATS Chi-X Europe used to be 10 times faster than traditional stock markets and was the first European platform to offer colocation in March 2007 in the Equinix data center in Slough, England. Other exchanges, such as the London Stock Exchange and NYSE Euronext, began offering colocation services in 2009–2010. More and more exchanges are now offering colocation services such as HitBTC, Coinbase, OKCoin and others.
While critics argue that colocation gives market participants an unfair technological advantage, Authority for the Financial Markets (AFM) argues that it is simply an investment decision that each market participant can make for themselves depending on its latency sensitivity, skills and consideration of the cost and benefits. The potential benefits that colocation can offer are offset by its high cost. In order to have fair competition, exchanges have been asked to ensure that multiple vendors are permitted for providing such services at their colocation facility.
Low-level programming languages
Like many other jobs, the advancement of technology is slowly shifting the skill sets needed for traders. As a trader, you will likely be interested in quantitative trading which includes high-frequency trading and algorithmic trading. It utilizes low-level programming languages to build the algorithms and ensure that hardware’s computing power is used efficiently.
Increasing automation can yield newer approaches that include more structure and more abstraction, allowing the guts of the programming languages to do what programmers used to have to do themselves. These automated features give the programmer more leverage and time to concentrate on the larger issues. In many cases, they also demonstrate better performance compared to the human decision-making process because these automated mechanisms are better able to find opportunities for efficiency and parallel computation while eliminating some of the simple mistakes that lead to errors.
Alternative data mining
While traditional data sources are abundant, alternative data provides a more rounded picture of the current state of the market. According to a 2018 survey from Dataminr, around 79% of institutional investors say they use some type of alternative data. Put simply, alternative data is information that has been collected from anywhere outside traditional data sources. In the financial industry, if the data comes from a source outside traditional areas like SEC filings, press releases or credit sources, it can safely be considered as alternative data.
Common alternative data sources include social media, web browsing behavior, cell phone geolocation, email receipts, credit card transactions, bill payments, and product reviews. Some use cases show how imaginative use of alternative data can lead to discovering important insights. For example, satellites that can count the number of cars in retailers’ parking lots or detect when manufacturers are adding shifts or reducing their workforces have garnered plenty of attention. In the commodity world, Genscape Inc., a unit of the Daily Mail & General Trust, uses helicopters to beam infrared signals at oil-storage tanks to gauge inventory levels ahead of government data. News reports also often focus on cloak-and-dagger-style data sources, such as the aircraft-tracking service that spotted a corporate jet owned by Occidental Petroleum Corp. at an Omaha airport in April, triggering hunches that executives of the oil giant were in talks with Warren Buffett’s Berkshire Hathaway Inc. Buffett indeed stepped up to the plate two days later with a $10 billion investment in Occidental.
Such hints can be crucial in making timely investing decisions and getting ahead of the market to acquire more profit. The use of alternative data is definitely the future of the financial industry. It’s more extensive, more up to date, and a better indicator than many of the traditional sources that have been around for years.
Neural networks and machine learning
Artificial neural networks are the basis of AI algorithms which are becoming increasingly common in our daily life. In machine learning, artificial neural networks form a family of statistical education models, created with biological neural networks in mind.
These are systems that are able to communicate messages to each other and have digital weight. This makes neural networks adaptable to input and capable of learning. That is why the systems based upon neural networks might be able to trade in and out without any repercussions by itself, as it would improve trading strategies based on the data it already received and processed.
Future intelligent systems could successfully implement all of the historical data from traditional and alternative sources that a human being is incapable of processing properly. It could employ technological advancements such as more powerful hardware and a combination of efficient programming languages to achieve a perfect performance and develop the most profitable trading strategy. It could be capable of checking out multiple market conditions across the globe simultaneously, saving a lot of time and eliminating any possibility of the slightest gap in time or occurrence of an error. It could eliminate major risks and market instabilities, with trading becoming more sentient and realizing the impact of a buy/sell gone wrong, to a dip in the exchange, rising up to the challenge to recover from that without any human intervention.
To put it simply, artificial intelligence could overcome those challenges and disadvantages in algorithmic trading that would not be possible for a human mind to do. However, it would be preferable if people and machines could work in tandem to achieve the perfect performance together. The team-work between humans and AI systems is really the key to having a market that is efficiently regulated while staying adequately transparent and fair.
The future looks promising because of the many ways technology can advance, affecting the entire market as well as each one of its participants. This is why every fresh idea has to be vigorously tested in natural market conditions and implemented carefully. It is important to implement technology that can interact with the market in a positive and efficient way so the advancements are beneficial for everyone.
Sources
Quant
AI
FPGA
High-Frequency Trading Acceleration
Colocation
Alternative data
Follow Us on Twitter Facebook Telegram
Featured Image: Shutterstock/Yurchanka Siarhei