One new direction of fintech is regarding robot programs that place automatically trades based on programs provided and certain parameters that the user chooses. My eye caught the following announcement on TechInAsia's website:
Robo-investor AlgoMerchant begins trading after $2m-plus funding
The Singapore-based startup offers a range of robo-traders that allow investors of all shades – from part-time retail investors to professionals and high-net-worth individuals – to automate securities trading through their personal trading accounts.
The robo-traders use data analytics and machine learning tech to automate trading, while also avoiding delays and human error. The basic service is free, while a range of premium packages can be paid for.
AlgoMerchant said it collaborates with freelance quantitative traders – in other words, those that specialize in automated trading – and data scientists from around the world to discover profitable investment algorithms. More than 1,000 traders tested out the service during its nine-month beta phase.
So far so good, it definitely sounds interesting.
But I almost dropped out of my chair when I read the following:
Forty percent returns
The startup claims that its bots can give everyday retail investors “an edge similar to resource-rich top quantitative hedge funds,” securing projected annual returns of over 40 percent.
Annual, consistent returns of over 40 percent are simply from another world. Even Warren Buffets returns would pale in comparison to those.
Just as an example, $ 10,000 would turn into $ 8,360,000 over 20 years using 40 percent returns. I guess we all would love that, it would for instance solve all pension problems of the world.
But worrisome, there is no basis whatsoever given in the article for these kind of returns, it looks like they are plucked from the sky.
On the company's website the only thing I can find regarding returns is this:
The majority of retail investors’ portfolios follow the returns of the market. AlgoMerchant’s strategies, however, are alpha-seeking and target upwards of 20% per annum.
Suddenly the "over 40 percent returns" has changed into "upwards of 20% per annum".
Following the same example, $ 10,000 would turn over 20 years into $ 383,000.
Not bad at all, but a rather far cry from the $ 8,360,000 based on 40% returns.
The graph supporting these claims is as follows:
- The starting point of any simulation is very important. In this case the company used January 1, 2008, in other words exactly at the start of the global recession. Thirteen months later equity markets are down 47%, the company claims Paladin (their algorithm) would theoretically be up by 45%, a huge outperformance by all means in a rather short while. But that crisis is a "one in a generation" event, it is very tricky to start a simulation exactly there.
- After the crisis, the central banks started the largest financial "experiment" (QE, quantitative easing) ever, driving interest rates down to a level not seen in 5,000 years. Again, the question is, how would Paladin have theoretically performed under more "normal" circumstances?
My guess is that these hypothetical results are derived from optimising on the data itself, causing over optimisation (especially given the rather unique circumstances of the last ten years), and thus generating much too rosy projections of future returns.