Continue in 2 seconds

Backtest provides the initial spark for this Frankfurt AI fund

  • Jan-Patrick Barnert
  • Bloomberg
  • March 05 2018, 6:30am EST

(Bloomberg) -- (Machine translation provided by Google and reviewed by Bloomberg editors.)

Columbus actually wanted to find a faster way to India and discovered America. The situation is similar for a German asset manager in the 21st century. For years, he has been trading successfully on the basis of an options strategy. Then he asks his IT administrator to test new strategies. Unplanned they develop an artificial intelligence that has recently become the manager of a new fund by the company.

The Frankfurt based asset manager Wallrich Wolf Asset Management AG has been in business with their short-options-fund for ten years. The portfolio managers regularly sell put options and earn the option premium. Until 2011 the fund was fully invested most of the time, which provided steady income, but did not allow scope for new positions when volatility increased.

Managing director Stefan Wallrich was keen to know, what strategies and optimizations work and asked his IT administrator, who is actually responsible for the maintenance of computers and servers, to support him in a so-called backtest. A backtest is a common practice in portfolio management to try new ideas or optimize existing ones. 20 strategies were calculated by Wallrich back to the year 2008. Among other things, this showed that, in principle, a variable degree of exposure in the portfolio, depending on the current market volatility, produces good results.

"Until then, I was happy with the results, because I just wanted to backtest some strategies to optimize our existing fund," says Wallrich, who manages more than 350 million euros ($427 million) with his team in various portfolios. But then came the big "enlightment".

Which degree of investment is the right one? Wallrich’s IT expert Marcel Heintz developed an artificial intelligence that constantly calculates different investment levels, checks them and questions the results. The machine is given the requirement to keep a portion of the portfolio permanently invested in order to generate regular income through the option sale and at the same time, depending on the level of volatility, to move the investment level up or down. With low volatility, less investment should be made and, with high volatility and thus high premiums, even more than 100 percent of the portfolio should be invested. Otherwise the machine is free to decided on the degree of investment by itself.

"The AI constantly assesses its results and questions how it came to the conclusion and how it can do even better in the future," explains Sebastian Franz, who was involved in the development as a DHBW-student.

Artificial intelligence technology differs from the more traditional quant investment approch, in that it mimics the human brain’s learning capacity. When new data comes in, the algorithm modifies its behavior. A machine detects non-linear patterns, without the help of a human being.

After a year of research and convincing results, Wallrich decides to give it a try and in December 2017 launches a new fund, which is managed only by the AI.

"We could have used the computer in our traditional strategy, but in order to give the topic the attention it deserves, we decided to launch a new fund," says Wallrich. "Also to distinguish the AI from the old fund and to be able to evaluate the success in the long term."

The fund currently manages 7 million euros ($8.5 million) and, like most AI funds, is still at the beginning. For Wallrich, the AI topic is not a revolution, but an evolution, and he wants to further improve the "intelligence factor" in the coming months.

For example, one could extend the AI by additional volatility variables or check how exchange rates or macroeconomic data change the decision path of the machine. In addition, it would be possible to launch several funds with basically the same strategy but different risk-return profiles. With the AI it is possible to manage several different portfolios cost-effectively and with less highly qualified portfolio managers.

In 2018 he aims for a fund volume of 15 million euros. In the medium term, "it can and should become significantly more," says Wallrich. However, that depends on the success of the AI strategy. "If it works, of which I’m very confident, the volume could sky rocket."

One has to go with the time and competition makes it necessary to stay innovative, Wallrich says. He could think of abandoning the classic portfolio management completely at some point and leave the money management fully to the machine. "The machine is simply better at many things, while we have emotional discussions in our portfolio management team," Wallrich says.

--With assistance from Bryn Colton

Bloomberg News