Blending styles, hybridization and the market mood “interpreters”: Quantamental and MetaQuant players
An everlasting debate: quant vs. fundamental philosophies.
Traditionally, fundamental and quantitative investing was treated as two separate modes, epitomized by two iconic figures: Buffet and Simons.
It is true that the fundamental approach does work well and brings in great results thanks to experienced analysts. But with the huge amount of data available today, not using it effectively means missing out on some great opportunities. One strategy seems not enough.
Therefore, in current times it might seem quite sensible to think in terms of blending the best of both these approaches human centered strategies and judgment with more data-driven tools in day-to-day decision-making to gain an edge. However, mixing them is easier said than done, as the two of them work in very different ways. There is certainly not a unique recipe. Besides, culture change is not easy for any organization. It is not a destination but a journey. Still we can stress the positive combining these two styles is a win-win. Why? Simply because it means:
(1) Bridge a gap: by merging advanced techniques within the fields of AI and ML with good judgment to monitor and adjusts the investment process to the prevailing market conditions.
(2) Avoid Cognitive bias: By employing a fundamentals-driven quantitative process with human oversight, investors can eliminate behavioral biases in an investment process.
(3) Timing: Quantitative techniques can streamline the investment process by effectively screening the universe for promising investments at a faster pace than possible by human analysts.
(4) Focus on added value: Technology can empower talented analysts to focus on ways they can add value to an investment process and let computers handle repetitive tasks.
(5) Improving performance: building a healthy culture which motivates two dissimilar groups to work together towards a common goal. They crux lies in creating a systematic process, organizational habits, and communication channels that value everyone equally.
(6) Balance: To round up, whether the quantamental investing strategy can generate consistent alpha depends on the institution’s ability to design a process that pairs data-driven professionals with traditional fundamental researchers. Finding a balance is necessary to create value from both methodologies.
Hybridization and guidance: a walk through the process.
In our own experience, fundamental analysis and quantitative techniques are symbiotic and strengthen each other at every step of the way.
It is true that hybridization of human expertise and computer power will augment the possibilities to gain an edge and in the best of the cases reduce costs. Still we believe an essential component is missing, a new holistic vision, a new player: the MetaQuant (as I define it: a profile which merges the linguist, the sociologist, the philosopher, the cognitive psychologists, the anthropologists) who will develop a synergetic environment reconciling the two styles.
We know that automation is not enough. Human judgment as added value is crucial when “translating, coding and interpreting insights into action”. Now, how do we accompany our customers through all phases of the process?
(1) Promoting a creative upfront thought: empowering more people with a diverse skill set to participate in working with data in an insightful way to look at datasets while developing common language to improve effectiveness and teamwork;
(2) More effective customer experience by designing use cases with the customer’s own data, training the staff, coding applications with novel technologies. Involvement is the key.
(3) From ideation to completion by designing an “Intelligent workflow” that involves a multidisciplinary team (Quantamental + Metaquants) to help customers smooth the shift through the full cycle process: Research, Code, Simulate and Automatize the AI based trading strategies.
(4) Expert Advisory to take cognizance about what is happening “under the hood” with their AI algorithms
- How to stay ahead of rapidly evolving AI models
- Scalability and budgeting to support advanced AI ecosystem department
- Gradual roadmap: plan, pilot, deploy
- Turning a “black box” into a “clear box” by resorting to explainable models
- “Interpretability models” a la carte
When the “narrative” needs blending strategies in the search of AI trustworthiness
With the advent of AI and its impact within the shifting markets, talent, alternative data, the conception of hybrid approaches such as the Quantamental and the MetaQuant have become key factors in a search for unexploded sources of alpha.
Having said that, one question that echoes in my mind is “Why should I trust your model?” Obvious as it seems, it is of paramount importance for the customer to understand what drives a model to take certain decisions and, of course, how to improve performance to gain an edge.
“Regardless how sophisticated explainable ML models might be, they can effectively be rendered powerless unless interpreted by human experts.” For us, interpretability is the upmost quality that machine learning methods should aim to achieve if they are to be successfully applied in practice. The higher the interpretability of a machine learning model, the easier will be for someone to comprehend why certain decisions or predictions have been made, mainly in a challenging environment.
Hence more, in many sectors, interpretability will be the catalyst for the adoption of machine learning and close the gap between those who claim that they do not use it since they cannot explain the models to others. We believe interpretability will address this issue and make machine learning attractive to organizations and people who demand transparency in how the model works.