Algorithm-based Investing: The Clairvoyant We've been Waiting for?🔗
It is no secret that elite angel investors chase statistical outliers. They are after those home runs that will become unicorns, generate an outsized return on investment, and make up for all the money lost on other startups in the portfolio that flamed out. No one, not even the angel investor himself, will remember the failed projects that never panned out, but everyone will be talking about the one or two startups that shot up through the roof in valuation.
Not many angel investors can find those home runs on a consistent basis, though. Those with a string of successful investments in their resumes describe their secret to success as a curious blend of gut feeling, lessons learned, and a healthy dose of risk tolerance. That's hardly a formula you can copy.
Like all of us, angel investors are limited by the amount of information they can digest and cognitive biases that can influence their decisions. But what if we outsourced investment decisions to artificial intelligence (AI)? Can technology offer help in devising a repeatable strategy to discover the best startups at an early stage? Should we do away with the old way of investment and turn to artificial intelligence and machine learning (ML) instead?
Cognitive biases: Enemies of rational thinking🔗
The proponents of algorithm-based investing emphasize how this method can help eliminate cognitive biases in order to justify its use. Angel investors, especially the inexperienced ones, are particularly prone to three common biases:
Local bias: Investors tend to mistrust and stay away from opportunities distant from their own environment due to the lack of information.
Overconfidence: Inexperienced angel investors tend to overestimate their knowledge and investment skills, trusting in their foresight in the face of a major decision. This assumption makes them ignore some of the available information, which undermines the quality of their decisions.
Loss aversion: While experienced angel investors see losing money as part of the game and chase home runs, inexperienced ones are more sensitive to potential losses. Worried that they have limited funds, these people tend to shun counterintuitive ideas. As a result, they miss out on some truly transformative ideas, which usually look counterintuitive at first.
ML to the rescue🔗
An ML algorithm is immune to cognitive biases and can make objective decisions while evaluating investment opportunities. This bias-free approach easily outperforms inexperienced angel investors hampered by cognitive biases.
According to a study researchers at the University of St. Gallen (Switzerland) conducted among 255 angel investors, the ML algorithm generated an internal rate of return (IRR) of 7.26 percent, while angel investors with higher levels of bias lost money at an IRR of -20.52 percent.
On the other hand, experienced angel investors, with an average IRR of 22.75 percent, proved more successful than the algorithm by a wide margin.
The key to leveraging ML for high-return investments is to find metrics that best predict startup success. AngelList has been one of the pioneers in this area, putting theory into practice and focusing on job applications a company receives in a quarter as their magic metric to track. The platform's job market arm, AngelList Talent, is used by around 2 million users every month. The AngelList team vets 35,000 companies seeking talent every quarter and narrows them down to half that number. Then, they reach out to 20 companies that are most popular with job seekers.
Abraham Othman, head of the investment committee and data science at AngelList Venture, believes that there is a correlation between hiring interest and company valuation and sounds confident in his company's investment model. Othman is particularly impressed by how the algorithm can cut through the noise like pitch deck clichés and investor bias and give the less-represented social groups a better chance to raise money.
However, in this specific case, the method's validity seems debatable at best. A company is rarely in complete control of the hiring demand it is receiving. Pay your employees more, and you will see more people willing to work for you, regardless of the potential of your product or idea. Or, if the rents are out of control in a particular location, companies located in places with similar living standards but lower rents will see a hike in demand, which has nothing to do with a startup's future potential. Scenarios like these cast some doubt on the method's viability, especially for angel investors chasing the next big idea.
The downsides of using ML in investment🔗
The limitations of algorithm-based investing are not metric-specific but rather structural. What comes out of AI is dependent on what you feed in. AI is good at pattern recognition, and it will get better at a task as it gets to train on bigger data sets. However, no matter how much data you feed into a computer, it will never be able to spot something like ingenuity or determination. Thus, there is no way for the computer to factor into equation parameters like the founders, their experience and determination, the team composition, etc.
The founder is the single most crucial factor for an angel investor to consider during an investment decision. The pitched idea may not look realistic at all, or the market may be nonexistent. But a founder who sounds like he has a plan and demonstrates the will to overcome everything to realize his dreams can tip the scales in his favor. Elite angel investors would not want to give up a chance to connect with such special people for the safe bets an ML algorithm would offer them.
So, where does that leave us? The Harvard Business Review study above argues that there is room for algorithm-based investment. The algorithm beats novice investors handily, but it is no match for elite angel investors. Therefore, training the ML on data based on investment decisions made by elite angel investors will make it sharper. This more advanced AI can, in turn, be deployed to help novice angel investors and shield them from cognitive biases that affect the inexperienced the most.
However, the weight of intangibles in successful entrepreneurship ensures that ML can only play a complementary role in investing. The ability to compose rules from patches of data is what sets elite angel investors apart from the rest. ML may guarantee mediocre results, but would Ron Conways and Sam Altmans of the world be content with mediocrity? Not likely.