Artificial Intelligence and the future of trading

October 31, 2018

What Elon Musk smoking pot tells us about Tesla's share price

A recent Economist article about the plight of the value investor got me thinking about the nature of stock markets and how they are today versus 20 years ago, and where they're going.

If a stock market is a distillation of pure capitalism, then as time progresses, the system optimizes and allocates capital (long and short positions) more efficiently.

25 years ago the process of trading stock markets was vastly different with open outcry still exclusively used on the trading floor of the NYSE until 1995.

Open Outcry trading

Low-frequency trading

The ebb and flow of the market and performance of individual stocks was more susceptible to the imperfect information and known biases of the individuals making the trading decisions.

Now a tiny percentage of transactions are private individuals buying a share for their retirement portfolio. The vast majority of transactions by volume and value are investment banks and hedgefunds who utilise vast computational clusters, huge streams of data, sentiment analysis and artificial intelligence to make microsecond decisions to trade.

How much of the value of individual stocks is now affected by human sentiment is unknown, but I view this as a positive change as the progress of capitalism is always towards increased efficiency.

Which leads me to consider what this implies for the future of trading and ultimately the valuation of companies.

Initially it would seem that these vast AI systems would learn to adapt to and optimise for human biases and prejudices. So when Elon Musk smokes pot on video, an automated trading system receives a flurry of tweets and determines the value of Tesla will drop and takes a short position.

Elon Musk smoking pot

Mulling the impact on Tesla's 3rd quarter results

The shock 3rd quarter profit and underlying fundamentals of Tesla are not affected, but the machine has learned how to interpret these signals as leading indicators of human actions and thus price impact.

As human sentiment plays a decreasing role in the impact of share prices (aside from the tiny fact that humans remain the customers, employees, shareholders, and so on, of these companies), trading systems will presumably make another step change in optimisation and remove the human element from all trading decisions.

That step change is in more and increasingly alternative data. This in not particularly insightful, given the business model of virtually every modern tech business is the accumulation of more data and the value ascribed. But the provision of alternative data and it's application to modern trading system is only the tip of the iceberg.

There have been some innovative steps such as using satellites to determine farm yields, image recognition of store carpark usage and sentiment analysis of tweets. But there's so much more to come.

As every aspect of our universe is digitized and categorised, increasing computational power and advanced machine learning capability means the system continues to optimise. Then the alpha that everyone is seeking becomes more, better and faster data.

That represents a longer-term trend. The interim step on this path towards increased efficiency of trading systems rests on improved fundamentals and more accurate derived data in guiding decisions.

Woodseer is one such data source, taking various data inputs and applying a forecast algorithm to produce the extremely accurate dividend estimates.

We're codifying the refined approach of our analysts into an algorithm and scaling that across over 20,000 stocks globally. The output is overseen by our analysts who are instrumental in the never-ending work of fine-tuning our algorithm to produce better results.

As always to learn more about our application of technology to produce accurate dividend forecast, please get in touch.