The only constant thing is that the edge is dynamic. There is no “code” you can crack to learn the markets.
When we are investing in the market, we are all looking for an ‘edge’. As the markets are a zero-sum game, an edge is an unfair advantage that we believe will land us on the right side of a decision.
This edge (whether real or perceived) can be your analysis of stocks, trusting the right expert, some intuition, technical analysis or using technology. The edge helps you be right a little more than 50 percent of the time, like a loaded coin in a coin toss that shows heads 51 out of 100 times.
The only constant thing is that the edge is dynamic. There is no “code” you can crack to learn the markets.
This is not true for other applications of AI – for example, training an AI system to learn how to drive a car, or perform surgery, all of these have some immutable laws, and an actual code to be cracked, i.e. the problem is solvable.
Good predictive models will all agree that an object before them is a pedestrian and not a tree or whether something is a tumour or a part of the body.
This is not true with the markets – five sophisticated models could look at the same data and come to a different conclusion for how to trade the next day.
The amount of news flow – political, economic, social that impacts markets daily is near impossible for humans to keep up with. In a single eight-hour trading day, there are 480 one-minute data sets.
The world is so increasingly complex and interconnected that even fundamental analysis, that which was once considered an “art”, is harder.
For example, blue-chip companies today are both stronger and weaker today. Stronger because of their competitive positioning and weaker because of ease of disruption today.
Therefore, the most effective edge is one that can constantly learn, adapt and change – enter AI and machine learning.
This is why the financial markets globally have been one of the earliest adopters of AI and machine learning. In 2001, IBM built a team of “robots” that beat humans at trading.
Sixty percent of all trading activity today is run on technology. In fact, the top five hedge funds in the world are all quantitative funds, using some form of technology to develop their trading strategies.
So, what do we know so far?
– Investing is zero-sum and we are all looking for an edge
– Markets run on data and humans cannot process this as quickly or efficiently as AI systems can – AI has an edge on humans
– Technology is running large parts of the global markets
Given all of this, how should you prepare for a world where AI runs the markets and eventually algorithms are all trying to “out-algo” each other?
Let’s break this up into two fundamentally different schools of thought on market participation – Trading and Investing.
Trading
“Artificial Intelligence is to trading what fire is to the cavemen,” as an industry analyst put it.
Most AI research, product, and strategies focus on trading strategies. This is the “lowest hanging fruit” of AI applications in the markets – process large amounts of data quickly to see if you can do a little more than 50 percent of your trades profitably.
High-frequency trading is of course fully ceded to machines – thousands of trades a second has no place for human involvement. Like we mentioned above, imagine in a single eight-hour trading day, analysing 480 one-minute pieces of data or 100,000 in a month.
Technical Analysis (TA) is a broader subject that involves studying patterns in price, volume, and other factors to take short-term positions.
Especially in India, TA is run by some combination of humans and machines depending on the frequency of trade and data looked at. However, this is increasingly a “solved” problem, and more and more of TA is run by AI.
Investing
Fundamental investing, i.e. studying a stock, its management, its moat, and future prospects to buy and hold has long been the domain of human investors.
Greats like Graham, Dodd and Buffett have been icons in the space that have helped build the body of work around this subject that has stayed constant for decades. In fact, Graham and Dodd’s seminal work, “Security Analysis”, was published in 1934 – 86 years ago! The premise is simple, fundamentals drive value, and this still holds today.
Therefore, at least in India, much of our equity investments (mostly through mutual funds, PMS, etc) rely on actual investing performed by human experts.
However, as we discussed above, the number of factors that impact a company has grown manifold since 1936, and humans are struggling to beat their benchmarks.
Eighty percent of large-cap mutual fund managers in India have been unable to beat their index over the last five years. This is because humans are emotional and suffer from biases, panic, and euphoria, no matter how disciplined.This is true everywhere. In 2008, Warren Buffett challenged the hedge fund industry to a million-dollar bet, and Protégé Partners accepted.
The bet was that over a 10-year period, a handpicked selection of five funds by Protégé (including fees and expenses) would not beat the S&P 500.
Throughout the bet, the S&P 500 returned 7.7 percent annually; the hedge fund basket averaged only 2.2 percent annually. Buffett won conclusively. This is an example of why in the US; more money is now managed by indexed funds and ETFs than active mutual funds.
Therefore, the question is, how can AI help? Can machine learning truly teach itself fundamental analysis? I believe the answer is, yes. If Graham was alive today or Buffett was just starting his career, I think they would both be big believers of using technology.
Warren Buffett has a quote in his introduction of “The Intelligent Investor” by Benjamin Graham: “To invest successfully over a lifetime does not require a stratospheric IQ, unusual business insights, or inside information. What’s needed is a sound intellectual framework for making decisions and the ability to keep emotions from corroding that framework.”
The best way to keep emotions out of investing is the use of technology.
We believe the next step in the evolution of AI in the markets is building algorithms that can learn and understand financial statements, industry movement, and global macros.
The premise is the same – to analyse millions of data points and gain an edge. The only difference from trading is that the edge is based on fundamentals and not just looking for price patterns in the markets.
Even in India, few funds have started building version 0.1 of this – using screeners and excel filters to apply some standard discipline to the investment process. This is a good start. However, those screeners are static. Static filters work till they don’t.
As we have discussed, the markets are dynamic and so your edge must continuously evolve. AI and machine learning will be version 1.0 – self-learning algorithms that can dynamically adapt to different market conditions and teach themselves what are the kind of fundamentals that are driving value in current market conditions.
It is not to say that AI or machine learning is fool proof. These systems are built and tested by humans, and therefore, there is always a worry of bias, data cleanliness, noise in the data, the robustness of the model to dynamically adapt, overfitting of data, etc.
Further, psychologically, there is a barrier to investing using technology because while a human can give you a rationale for the stock, how will a machine do it? The opposite side of that argument is that human brains are also black boxes!
Sure, they can explain their logic, but a lot of the time the logic has a percentage of “gut” or “intuition” involved. With a machine, at least the process it will follow will be consistent even if you cannot always understand the code that drives it.
To overcome this as an investor deciding whether to back a tech product, it will ironically come down to your reading of the people behind the machine. Their approach and logic to investing are what will show up in the machine.
Therefore, for you as an investor, how do you use AI to invest?
Step 1 is always asset allocation. Decide your allocation to equities and then within that diversify the approaches you invest behind. The definition of an approach can be a style (value/ growth), market caps (large/ mid/ small), passive or active funds, and tech or human. Conventional wisdom will dictate that you diversify.
Diversify your approaches and, of course, diversify away from humans by allocating to tech products as well.
Therefore, in conclusion, ask yourself this: do you think 10 years from now; there will be more technology in investing or less? If your answer is more, you need to start thinking today about how to get it to work for you.
Disclaimer: The views and investment tips expressed by experts on Moneycontrol.com are their own and not those of the website or its management. Moneycontrol.com advises users to check with certified experts before taking any investment decisions.
