Systematic or quant investing refers to using AI for asset management. Quant funds use AI-based quantitative models for alpha generation. And also, for other asset management tasks. Investment firms use AI to predict stock prices, create portfolios, execute trades and manage risks. They also use AI for operational tasks and regulatory compliance. Other uses of AI include customer acquisition, cross-selling to customers, identifying customers at risk and customer support.
What is systematic investing?
Systematic or quantitative (or quant) investing refers to using data-driven approaches to the entire gamut of asset management activities. Asset managers in quant funds or systematic funds use machine learning and AI for predicting asset prices, creating and rebalancing portfolios. Also, they use alternative data, in addition to the traditional time series data, company financials, macro-economic data and industry-level information.
Alternative data includes text, image and video data from different sources such as company statements, analyst reports and social media posts. It may also include data that has no obvious connection to the asset in question such as container traffic, rainfall data, export and import data and so on.
Systematic funds also use AI for automating operational tasks and regulatory compliance. And most funds also use AI for customer acquisition, customer support as well as for cross-selling additional products.
Man Vs Machines: The difference between discretionary investing and quant investing
For decades, discretionary investing has been the norm. Usually, discretionary fund managers are investment wizards with years of experience. They have a particular view of the market and they rely on their gut feeling, instincts or personal judgement. Of course, they use fundamental analysis and leverage their deep knowledge of the industry. They usually bet big on certain companies and to be fair, many of them have been quite successful in the past.
Quant investors are a different breed altogether. They don’t believe in gut feel or hunches. They use hard data to make investment decisions. They remove human emotions from their decisions. They use quantitative models to test out their hypotheses and back test all their strategies. And they use technology to execute trades efficiently in multiple markets and deal in several assets.
Quant investors also use machine learning and AI algorithms to create portfolios and rebalance them. And most systematic funds use technology to manage all the supporting functions like operations and customer support and so on.
Rather in support of this approach, Daniel Kahneman said, “Experts are uniformly inferior to algorithms in every domain that has a significant degree of uncertainty or unpredictability, ranging from deciding winners in football games to predicting longevity of cancer patients. One has to accept financial markets are no exception to the rule.”
The discretionary investing idea looks quaint today because new age companies are born at the intersection of multiple technologies and the markets are dynamic and the investors are impatient. So, discretionary investing is going to be a thing of the past.
What do asset managers use AI for?
Quant fund managers use AI for a variety of tasks. You can use AI for predicting asset prices and executing trades efficiently. And for creating and rebalancing portfolios. Also, streamlining operations, managing risk across portfolios and regulatory reporting and compliance are tasks that AI can easily handle. Other important functions that AI can do well are customer acquisition, onboarding, customer support and even cross-selling other products to improve profitability.
Let’s look at each of these areas closely.
Alpha generation using AIGenerating alpha is the goal of any fund manager. What is alpha? Alpha, in the context of investments, refers to the excess returns that an investment strategy makes over a benchmark. So, alpha is an indicator of performance. To generate higher alpha, you need to pick the right assets that will perform better than the market or benchmark indexes. And also reduce the cost of operations. Quant funds use advanced machine learning and AI for generating alpha. Also, to reduce total costs.
Systematic funds use machine learning and AI for their core investment processes such as identifying assets that have above-normal appreciation possibilities. Also carry out automated trading in a variety of assets and in several exchanges. And create portfolios optimized for high performance as well as rebalance them automatically.
With the rise of deep learning algorithms, making sense of complex data such as text in news articles, blog posts, statements issued by company executives, industry leaders and analysts has become easy.
But what’s deep learning? Deep learning algorithms try to mimic how our brains function. We know from our high school biology classes that our brains have billions of interconnected neurons. Deep neural networks use layers of artificial neurons. And this artificial neural network (ANN) can quickly make sense of complex data and create insights impossible for human analysts to come up with. A branch of deep learning called NLP (Natural language processing) can process text. So, with deep learning, you can process a huge amount of unstructured text data like social media posts such as Facebook posts and tweets. And on real-time!
Also, deep learning algorithms automatically identify the relationships between independent variables (such as raw material availability) and the asset prices. This frees the asset managers from coming up with hypotheses about these relationships.
These modern methods help asset managers leverage what’s known as ‘alternative data’. Alternative data is information other than conventional data such as time series data, company fundamentals and macro-economic indicators. So, alternative data includes variables that are seemingly not connected to asset prices. For instance, data such as container traffic, satellite data about crop density and even weather pattens are alternative data. It also includes text data from social media etc.
You can also use algorithmic trading software to buy and sell several assets in different markets at the same time. And you can even split the trades so that your trades don’t affect market prices. Also, you can execute these trades automatically without your emotions getting in the way. Finally, the software takes care of all the administrative stuff such as settlements.
Using AI for risk management and regulatory compliancePost the 2008 financial crisis, most governments and central banks across the world started tightening regulations for financial services companies. Also, fraudulent transactions, money laundering and privacy breaches have skyrocketed in the recent years. So, regulatory reporting and compliance have become major concerns for investment firms. There are multiple agencies that frequently update regulatory requirements. As a result, fund managers need to send multiple reports to different agencies. And the penalties for non-compliance are quite stringent. Huge fines and even jail terms in extreme cases! This has led to the rise of a new class of RegTech companies that use emerging technologies to make regulatory compliance swift and easy.
Quant funds use these AI-based tools for KYC (Know Your Customer), AML (Anti-money Laundering) and CFT (Counter Financing of Terrorism). And to report suspicious transactions, identify any fraudulent transactions and comply with authorities on regulatory reporting.
AI for customer acquisition and related activitiesYou can use machine learning and AI algorithms for customer acquisition. For example, unsupervised algorithms such as clustering, can segment a large customer base into micro-segments. This method groups customers based on their similarities to one another. You can then customize your marketing mix to each of these segments.
Quant funds use linear regression to calculate the LTV (life time value) of individual customers. Linear regression is a machine learning algorithm that uses past data to predict a continuous value. Examples could be the price of an asset or the earnings of a company.
You can also use AI-based recommendation systems to suggest more products for customers. Just like how Amazon and Netflix do. Many funds also use AI to identify customers at risk- customers who are likely to leave the fund.
With the growth of deep-learning based, NLP (natural language processing) algorithms, chatbots can communicate with customers to handle their routine questions. You can use AI-based content writers to create content for blog articles, email subject lines, email messages and even social media posts. Sentiment analysis tools can help you understand your customers’ attitudes towards your brand. And take corrective actions.
Using AI for operational tasksAsset management firms have to deal with customer emails, requests for information, regulatory reporting and internal management reporting. Most of these are repetitive tasks and hence can be easily automated with RPA (Robotic Process Automation).
What is RPA? RPA refers to a set of tools called bots, short for robots, that will do all the routine, repetitive stuff that human staff usually do. The advantage of using RPA is that you can get a lot done in a short amount of time and also gain insights into how to improve the processes over time.
Also, AI-based HR analytics software can automate staff recruitment and performance appraisal. And take care of salary and benefits administration.
What are the advantages of using AI in asset management?
- AI removes the risk arising from human emotions and judgment errors. All decisions are data-driven.
- Better alpha generation. Using alternative data from a variety of sources can improve alpha generation. And this is true even when you are unaware of how the variables affect the asset prices.
- Using AI based technologies across several functions in asset management increases efficiencies. And also slashes costs across the board.
- AI can help you get new customers, keep them and serve them in a cost-effective manner. And also help you cross-sell more products to them and increase profitability.
Arguments against quant investing
- Argument: Financial markets are dynamic. And because of constant trading, the prices keep changing. How can you use machine learning to predict the future prices in this scenario? Machine learning algorithms can handle real-time data that comes in at high velocity. In fact, prices that change every second is additional data and can be useful for prediction.
- Argument: Lack of availability of a large amount of data, especially about an asset that has little time series data. This is probably true of an individual asset. But quant funds use a lot of alternate data. And that could be container traffic, satellite pictures of crops and text data from web sites, social media, company statements and media discussions. Together, we have a lot of data required for the model to predict prices.
- Argument: AI models are opaque and systematic funds are hard to understand.There is some truth in this argument, especially when we use deep learning models. Deep learning models learn recursively. And they arrive at the weights for the input variables by comparing their estimates to the actual prices. So, we usually have difficulty explaining why the model arrived at the price it calculated. Data scientists are working on what’s known as XAI or explainable AI. And that can show how the algorithm arrived at the result that it did. But this doesn’t have any negative impact on the performance.
- Argument: Since the upfront cost of technology is high, quant investing results in lower benefit-cost ratio. This is a weak argument, at least in the long run. Most of the set-up costs are one-time costs and the benefits add up over a period of time. Most quant funds use a lot of open-source software that are free. Also, you can tweak the models and re-use them several times for different assets.
- Argument: Limited amount of data on an asset will lead to overfitted models. This is a real danger when you use a limited amount of data and the model may not perform very well when exposed to completely new elements in the market. But, again, using alternative data and using a lot of test data can train the model well. Also, you can tweak the model based on performance.
- Argument: Execution of trades based on AI models needs constant supervision. This is true and you need to track all trades and ensure that the software and hardware don’t malfunction.
What is the future of systematic investing?
AI models will become quite easy to use, because cloud majors such as AWS, Google and MS Azure are working on democratizing AI. You’ll be able to develop, test and deploy sophisticated models with just your mobile phone. Also, AI will become more explainable. Plus, researchers are working on predictive models that can work on ‘small data’.
New startups at the forefront of disruptive innovations are working on the convergence of emerging technologies. For example, a stock like Tesla represents the convergence of many domains such as battery technology, autonomous mobility, AI and related technologies.
Another example is Atomwise, a synthetic biology company working at the intersection of biology and AI. Analyzing stocks like these requires knowledge of genomics, AI and health care domains. The point is that your conventional industry analyst will not be able to make sense of data from many domains. You need cross-functional teams that are also connected to experts and investor communities on social media platforms. And this approach generates huge volume alternative data and only quant models can process them. This is an approach espoused by Cathie Wood of Ark Invest.
The alternative data will grow with the expansion of the population using the Internet and social media platforms. There will also be a lot of vendors selling third party data.
All things considered, the asset managers with the best algorithms will win!