The need for data

The total amount of data created, captured, copied, and consumed in the world is forecast to increase rapidly. The development of digitalization contributes to the ever-growing global data sphere. There is a case to be made that the quality of data and how it is used has never been more important for fund managers. Using it effectively helps to analyse and predict market patterns ahead of time and consequently adjust risk and build investment resilience accordingly, which can provide a priceless competitive edge. Furthermore, despite the proliferation of the data, relevance is not guaranteed. 

Yet, a recent Deloitte report found that a just a small percentage of financial analyst reports mentioned Covid-19 between 1st January and 6th March, despite the virus spreading in Europe from February. Even amongst the reports which referenced the virus, only 28% were negative.¹ It was clear that, with the data they had, many analysts did not see the outbreak causing stocks to plummet. Applying advanced analytics, using Artificial Intelligence techniques, to these data sets may have provided more real-time insights into the risks associated with Covid-19 for the stock market. 

However, fund managers were compromised in their ability to react to the severe market changes brought on by the Covid-19 pandemic through their use of traditional risk management models. Built around strong assumptions on the behaviour of underlying assets and measuring normal distribution patterns on linear scales, these models are effective during times of relative stability. 

However, this also means traditional quantitative models cannot handle changes to the entire framework of the market. As a result, in non-linear periods, where black-swan events such as the pandemic can turn the markets on their head, traditional quantitative models aren’t able to accurately analyse or predict where the market will go next as floods of chaotic data pour in. During the pandemic, fund managers had difficulty deciding how to navigate their way through the crisis, manage risk effectively and make sure investments were protected. 

Deep impact

Advanced AI models may not be able to predict the unpredictable but they can detect anomalies well ahead of time and serve as a powerful tool to anticipate future market events that are considered unique, such as the Covid-19 pandemic.

For example, in February 2020, Axyon AI’s anomaly detection system signalled that the oil market was behaving anomalously, reaching an unprecedented level of 100% anomaly. Three weeks later, Saudi Arabia initiated a price war with Russia, contributing to a 65% quarterly fall in the price of oil, alongside the impact of Covid-19. While Axyon AI’s model didn’t predict the price war, it did detect the anomalous data and that markets were highly unpredictable. An investor benefitting from such information, could have reduced their exposure to oil, for example, and waited for more predictable times.

Rather than rely on linear patterns as many traditional models do, advanced AI systems remain completely agnostic, meaning they can be trained just to model what is in the data, with no theory behind it. At the same time, when the structure in the data shows an anomaly, the AI is trained to raise the alarm about the possibility of an upcoming unpredictable event, giving fund managers more time to prepare and adapt.

Out of all the different types of leading AI technology, deep learning is the one that fund managers could gain the most from. Through deep learning, fund managers can capture complex, non-linear patterns in asset behaviour and allow algorithms to adapt to changing market conditions continuously. Implementing deep learning would remove concerns about whether a model could cope with chaotic data flooding in.

Additionally, deep learning can also help fund managers unlock expanded uses of alternative data, a field which is rapidly growing. Alternative data goes beyond traditional sources of information like government statistics and incorporates sources such as social media, online searches, and credit-card transaction data to gain new insights. It’s projected that investors will spend $900 million on alternative data by 2021 globally, nearly double the 2017 level², yet many will struggle with being able to utilise the data effectively without the superior analytical power of AI and deep learning.

Although deep learning and other AI-powered tools could be used by fund managers to enhance current strategies, using them to build a new one from scratch could have the desired impact. Using a new strategy, totally AI-powered, fund managers diversify alpha generation in their investment process and allow to mitigate risk of the entire portfolio.

Results are being seen in those funds which have built artificial intelligence capabilities with AI-led hedge funds producing cumulative returns of 34%, compared to a 12% gain for the global hedge fund industry between May 2017 – May 2020³. The competitive edge that AI-powered tools can provide is already driving a shift towards the greater adoption of these tools, with 56% of hedge funds stating they would be using machine learning in their trading process by 2021⁴.

Fresh approach

Despite the challenges presented by the current situation, it’s been an opportunity for fund managers to reflect on current investment strategies and whether the provisions in place are up to scratch. Traditional risk models have been shown to be unable to model a landscape with a sequence of extreme events.

Moving forward with an AI-powered model, such as deep learning, can provide a flexible, powerful, and insightful solution to fund managers’ problems. They could give managers the extra time needed to implement more effective risk protections, against a market crash while continue to generate alpha through their superior predictive and analytical power, despite the current uncertainty.