By Alejandra M.J. Litterio1
1Co-Founder and Chief Research Officer, Eye Capital Ltd
With the advent of artificial intelligence (AI) and its impact in capital markets, the analysis of traditional data – such as historical publicly reported revenues, fundamental, technical and economic indicators – seems to be no longer enough.
Talent, alternative data, the conception of hybrid forms such as the “quantamental”1 and the “MetaQuant” have become key factors in a search for unexplored sources of alpha. How do we develop a comprehensible model with the most groundbreaking strategies based on machine learning (ML) and AI that allows investors to reap greater gains than other market participants?
In this chapter we rethink the core value proposition of alternative data and why a MetaQuant approach is a meaningful advantage to investors, enabling a deeper understanding and correlation of all the sources that could improve the appraisal of alternative data as a key engine for innovation in the “arms race” for an edge.
Historically, in almost any field, there is, apparently, a “holy grail”. In finance, it is the quest for novel sources of information as a way of gaining an advantage over competitors, called alternative data (alt-data). As alt-data continues to enter the mainstream, becoming an essential part of portfolio construction, investment professionals, hedge funds and asset managers have begun to conceive it as the most precious resource changing the capital markets landscape forever.
Of course, those who do not follow this seismic shift and update their investment processes accordingly face an increasing risk of lagging behind. However, finding an edge does not only depend on the quantity of data, but also on the ability to combine data sources that otherwise seem unrelated, and encompassing data to extract insights from non-financial and non-traditional sources to improve alpha generation.
Defined as “…any information collected from non-traditional sources used for a different purpose than the one initially intended”, most conceive alt-data as a niche of unstructured unorthodox data not previously used in the investment process. For others, alt-data is simply any information that is non-market data. Some of the most common examples include geolocation, satellite imagery, apps, social networks, IOT (sensors), microdata about consumer shopping (credit card transactions), and insurance data. But the definition is much broader, and clearly brings us somehow closer to visualizing a wider field of effective action to create better predictive models.
Because of the diversity and high volume of alt-data sets, a new issue has arisen: how alt-data is characterized to determine which one is most valuable according to the organization’s needs. In this respect there is not a unified or formal criterion to discriminate between conventional or unconventional alt-data. Kolanovic and Krishnamachari (2017)2 have presented a taxonomical approach based on the usefulness of the alt-data set. Dannemiller and Kataria (2017)3 proposed a continuum from structure to unstructured data. Other authors have adapted Kitchin’s system (2015)4 and designed a six-dimensional model. No matter how alt-data is classified, though, the truth is that market participants must think about:
Are traditional data science teams good enough to develop models able to achieve long lasting competitive advantage? Not quite. The complexity of incorporating unstructured data into the portfolio construction process makes it necessary to create a new profile of quants. We have coined the term “MetaQuant”.
The MetaQuant is a new breed of market player who is able to “translate human language into signals” and “read” the data from a holistic perspective. They are linguists, semiologists and philosophers – or rather a combination of these three intertwined profiles – who will fuel the potential for information advantage and provide a unique core differentiator, transforming data into knowledge in the financial world.
The MetaQuant has emerged as a crucial component of any AI quantitative trading model and paves the way for a novel insight: the “reinterpretation of alternative data”. It is in this scenario where the combination of structured and unstructured data using AI technology in a big data environment has a hidden value.
In the common quantamental process, advanced computational techniques are used extensively to reduce bias and random noise in investing. The MetaQuant takes this thinking “out of the box”, addressing crucial financial questions from a more descriptive qualitative level, integrating the unstructured with existing models, or creating a differentiated model.
The MetaQuant, based on a holistic approach, explores unstructured data within a specific context disclosing in-depth combinations of multiple factors from a variety of data sets providing valuable insights that are not so visible to the “naked eye”. Once the corpora (unstructured +alt-data) has been exhaustively analysed, the MetaQuant will design an integrated framework with variables defining all the possible relationships (for instance, finding patterns extrapolating and identifying phrases highlighted in an earnings conference call associated with geodata or even market drivers) which will be modelled by sophisticated AI algorithms, fuelling and enriching the portfolio selection and its rebalancing while assigning a predictive value or score to each asset allocated by the portfolio manager. With a hybrid model, portfolio managers will have at their disposal not only an incredible wealth of data and disruptive analytics tools, but also potentially new distinctive and actionable insights in the search for an edge.
The landscape of data is ever-changing and, as a consequence, market participants need to evolve to stay ahead in order to gain a unique competitive advantage and boost profits.
This is just one piece in the mosaic of the financial puzzle. While still in its early days, alt-data is a (r)evolutionary step towards bringing transparency into capital markets and global economies. The MetaQuant is certainly the figure “turning over the most rocks” at an infinitely greater scale, critically questioning every stage of the model, monitoring and revisiting the interactive framework to give investors an edge.