Opportunities for Quant Trading
What opportunities are there for systematic trading? Where might it go next?
There are a ton of opportunities for Quant Trading in the future, too many to list here. Some fraught with their own unique problems. Here are some of my thoughts based on my knowledge and understanding of space.
I’d love to hear from those in the sector what they think. Hopefully, this also gives juniors and those on the outside looking an understanding of where the sector is heading.
Opportunity One
Machine learning (ML) has created tremendous opportunities for systematic trading, a rules-based approach to trading. The trading system has improved since the 1920s when traders predicted moves based on charts, manual calculations, and hand-drawn graphs on glass. Now, trading systems use better computer processing and additional data. This new equipment has enabled traders to automate the process, allowing for quicker calculations and a larger scope of research. However, the process cannot be entirely machine-run. Humans are still needed during the research process to build models and use ML to predict signals. Logically, the next evolutionary step is teaching the system to perform its own research and to create its own signals.
Machine learning (ML) would improve all three key areas where algos are used: execution, prediction, and risk management. In execution, ML could effectively process order book data on a microstructure level. Machine learning could also more accurately predict alpha by incorporating more data into the process, particularly alternative data. It could also improve risk management by discovering correlations that enhance hedging. This benefit is essential to avoid past failures such as the bear market’s shut down in the 1970s when 170 out of 200 hedge funds were lost due to incorrect protection.
I wrote more on Machine Learning and its likely impact on quant trading over the next ten years here.
Despite its benefits, machine learning is not without issues. Successful systematic trading, especially machine learning, requires data. However, data and infrastructure can be tremendously expensive. According to AlternativeData.org, “Total spending on alternative data by buy-side firms will jump from $232 million in 2016 to $1.7 billion . . . in 2020.”[1] This prediction threatens high expenses to those involved in systematic trading. Even if one is not spending on data, they will be overtaken and beaten by those who are using it effectively.
Another issue involved in machine learning is uncertainty about how its models work. Machine learning models are more complex than quant models and more difficult to learn. With quant models, one can essentially “understand” how their model works. However, this is not always the case with machine learning. A ML model may perform a task for an unknown reason. Not being able to debug it is an even bigger problem than understanding. Beyond this, forming relationships with investors and effectively articulating quant functions are keys to successful fundraising. Explaining how ML models work will be crucial to investor confidence, capital raising, and more. Explainable AI is a hot topic for this reason.
Quants should be cautious when using ML because machine learning can find patterns that do not exist. A famous example is the Anne Hathaway Effect caused by news-reading algos. Many machine-learning algos were “reading” news articles and tweets that included the word “Hathaway” and deduced that Berkshire Hathaway stock was popular. However, the news was referring to the actress Anne Hathaway and her recent success. The mix-up caused the ML sentiment algos to make Berkshire Hathaway stock go up whenever Anne Hathaway was in the news! [2]
Another item one must watch for is biases. An advantage of systematic trading, and arguably its greatest strength, is the removal of human biases. However, a known concern is unwittingly incorporating biases into ML algos through the datasets used.[3] For example, the algorithm that Amazon used between 2014 and 2017 to screen job applicants reportedly penalized words such as “women” or the names of women’s colleges.[4] Ensuring one does not incorporate biases has always been hugely important, especially now in the current cultural climate. Including biases in algos that exploit biases would be ironic.
Opportunity Two
A second opportunity for quant trading is its ability to grow into new areas. The two hottest areas are fixed income and crypto. Systematic trading has always evolved into new markets and new strategies. Famous firms such as AHL started by following trends, and now it trades multiple assets in a multi-strategy approach. Jump Trading began as an HFT prop shop, but now it trades multiple assets in multiple frequencies and has a VC arm. Nearly all quant funds expand and seek new sources of alpha. The decay of signals has accelerated in modern times. Previously, a signal may have lasted for weeks, months, or sometimes longer. Now, the competition quickly finds them and arbitrages them away. Because of this struggle, finding new alpha is a constant battle for quants.
There are basically two ways to approach the problem: systematize an existing strategy or take one’s systematic model to a new asset. A quantitative, systematic, data-based approach is preferable to one that is purely discretionary. For example, an event-driven data analysis successfully predicted a merger arbitrage opportunity by looking at flight tracking data.[5] Quants saw a company jet flying to a particular airport in a seemingly random location and successfully deduced the company was flying there to meet with another about a proposed takeover.
Quantamental has been a buzzword for a few years now where quantitative and fundamental worlds collide. This space is growing by the day as more fundamental shops look to incorporate data and quantitative insights into their investment processes. Typically, this combination takes two forms. At the start, quants rank companies quickly and on a massive scale, allowing the analysts to focus on the best. Quants also work with the analysts to supplement their fundamental deep-dive research work with additional insight from alternative data sets, such as credit cards, foot-fall traffic, or perhaps satellite imagery.
Systematic trading could also expand into new asset classes in illiquid markets which have historically been reserved for manual execution, as eluded to above. Here, it would benefit from automation in algo execution. Fixed income, including credit (and particularly corporate credit), is being increasingly electronified. Years ago, equities experienced this process when the fixed income world had been out of reach. But, as opportunities become harder to find in equities and the sell side improves its ability to offer electronic trading, opportunities arise for systematic trading in the fixed income and credit worlds, especially while inefficiencies remain. There are approximately 41,000 stocks but millions of differing bonds.
Cryptocurrencies are (and will be) a big area for growth. The more markets that are traded systematically, the harder it will be to find inefficiency and make profits. One option to improve this method is to move on to inefficient markets such as newly electronified fixed income or volatile new crypto markets. Many groups are already operating new strategies and are actively hiring. Paul Tudor Jones of Tudor Capital began investing a while back.[6] BlackRock has approved the trading of bitcoin futures.[7] DRW has had a crypto business for a while now. Jump is heavily invested in the space. Steve Cohen, owner of Point72 has recently been convinced.[8] Just to name a few.
Systematic trading needs new spaces of opportunity because of a self-fulling prophecy and paradox. Systematic trading works by exploiting market inefficiencies. But, as it finds and exploits the inefficiency, it improves efficiency and reduces opportunities to exploit. The 2007 quant crash was the result of overcrowding and too many algos unwinding at the same time, showing that too many similar algos can harm the market. So, in essence and in irony, as systematic trading makes the market more efficient, it reduces its own ability to perform. Ergo, systematic trading needs to continue to move on and discover new inefficient markets to maintain its success, especially as more automated trading removes human biases.
The other solution to the systematic trading efficiency paradox and the third way Hedge Funds will grow in this area is to improve current processes. By having better data, better infrastructure, better execution, and better talent than their peers, businesses can aim to exploit inefficiencies by “seeing” them first either through speed or knowledge. Going forward, hedge funds must seek profit through weaker and more nuanced signals.
Where do you see opportunities for systematic trading?
[1] https://www.marketwatch.com/story/the-explosion-of-alternative-data-gives-regular-investors-access-to-tools-previously-employed-only-by-hedge-funds-2019-09-05
[2] https://www.theatlantic.com/technology/archive/2011/03/does-anne-hathaway-news-drive-berkshire-hathaways-stock/72661/
[3] https://towardsdatascience.com/biases-in-machine-learning-61186da78591
[4] https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G
[5] https://www.marketwatch.com/story/the-explosion-of-alternative-data-gives-regular-investors-access-to-tools-previously-employed-only-by-hedge-funds-2019-09-05
[6] https://www.cnbc.com/2020/10/22/-paul-tudor-jones-says-he-likes-bitcoin-even-more-now-rally-still-in-the-first-inning.html
[7] https://www.bloomberg.com/news/articles/2021-01-20/blackrock-files-to-add-bitcoin-futures-to-two-of-its-funds
[8] https://www.bloomberg.com/news/articles/2021-09-15/steve-cohen-throws-himself-into-crypto-after-early-skepticism
