Q2-2021 EXAM: EARLY REGISTRATION OPENS ON DECEMBER 7, 2020

Resources

Events  | News | Resources | Webinars |

WHAT DO RACE CAR DRIVING AND CAPITAL MARKETS HAVE IN COMMON?


August 4, 2020 
By Mehrzad Mahdavi, Ph.D., Executive Director, FDP Institute 

Let's look at the recent headlines1,2

  • Next year will be the first-time race cars cross the finish line at the Indianapolis Motor Speedway on their own, no drivers!
  • Some recreational flyers can now buy private planes that can safely land themselves in emergencies, no pilots needed!

Indeed, significant milestones are pointing to paradigm shifts in transportation. However, the real stories are more about the next generation of AI, including real-time artificial neural nets (ANN) and cognitive learning.

And, this is where the race car driving industry and financial services intersect! 

Handling the race conditions will push the envelope of what is possible with AI in autonomous vehicles. For the first time, the AI systems will be subjected to the pressures of professional racing conditions, with speeds of up to 200 miles an hour and the need for split-second decision-making around collisions. They also will have to deal with factors such as wind shear and slipstream physics.  These are the ultimate stress tests for AI performance in real-time.

So, how do these advancements in AI provide benefit in the Financial Services?   

Artificial Neural Networks have a clear use case in trading execution3.  The real-time aspect of AI being [stress] tested in race-car driving is going to be very helpful in building robust models for high-frequency financial strategies.  Applications to hedging have also been referenced4.  Other applications such as portfolio management have been eloquently described by Ganesh Mani: "Self-driving investment vehicles - that automatically adjust allocations - is an exciting new frontier for both individual and institutional pension portfolios. Event-driven rebalancing, keeping in mind the portfolio persona, reacting to new - traditional as well as alternative - data and market events can add value to and subtract emotions from the process mix!"

When it comes to the application of ANN in financial services, differences with autonomous driving should be noted as well.  In autonomous driving, there is a code to be cracked. The problem primarily involves geometry, laws of motion, and road maps — all stationary items.  There are also large training data sets thanks to many vehicles operating under real-world conditions.  

Is there really a "Code to Crack" in the financial markets by using big data and machine learning5,6?  There are companies that are diligently working on this problem.  An example is a tech company – not a financial firm – using deep learning models and neural networks for trading7.  They suggest the possibility of beating the stock market is no longer theoretical.  And, they have some excellent results to support the claims.

However, other industry experts argue that the role of artificial intelligence in financial markets 'isn't to find the "Holy Grail" or "Crack the Code", but to have processes that can recognize changing conditions and opportunities and adapt accordingly.   This approach is driven by specific characteristics of financial markets that is different from autonomous driving.  First, financial markets change all the time, driven by political, social, economic or natural events; Second, any new insight or "edge" is copied quickly and competed away; Third, financial data are often sparse and more nuanced compared to autonomous driving.  Therefore, techniques such as Exploratory Data Analysis (EDA) and "training on the tail" become essential analysis tools. 

In short, financial services gain quite a bit of advantage from autonomous race car driving with AI at its core.  However, as usual, the devil is in the details!

Key takeaways:

  • Mind the data – artificial neural networks work with large training data sets (as in autonomous driving).  Use cases in capital markets require a more nuanced treatment of data, including EDA, training on the tail, alternative data, etc.  
  • -Follow well-specified, repeatable processes to ensure consistent use of scientific methods.   This methodology allows you to identify new regimes and to make adjustments in training data and models.
  • Have an in-depth understanding of inherent uncertainty around the models and the range of performance outcomes you should expect.

Final Thought: "AI and machine learning is an interdisciplinary field.  The diversity of industries and applications is a critical factor in lifting the capabilities for all"8.

References:

  1. https://www.wsj.com/articles/press-a-button-and-this-plane-lands-itself-11595044800?mod=djemAIPro
  2. https://www.wsj.com/articles/autonomous-vehicles-to-race-at-indianapolis-motor-speedway-11595237401?mod=djemAIPro  
  3. Conversations with Tony Guida, Advisory Board, FDP
  4. Conversations with Nigel Noyes, Advisory Board, FDP
  5. https://www.marketwatch.com/story/machine-learning-wont-crack-the-stock-market-but-heres-when-investors-should-trust-ai-2020-06-08
  6. https://towardsdatascience.com/predicting-short-term-stock-movements-with-quantitative-finance-and-machine-learning-in-python-e6e04e3e0337
  7. https://www.fastcompany.com/90502428/artificial-intelligence-beat-the-stock-market
  8. https://bdtechtalks.com/2020/07/13/ai-barrier-meaning-understanding/






  Brought to you by:

 

2020 Financial Data Professional Institute©

Contact Us: info@FDPinstitute.org 

Privacy | Policies | Terms of Use | FAQ

Powered by Wild Apricot Membership Software