Stockholm (HedgeNordic) – Established in late 2012, Minastir AI Currency Fund is an artificial intelligence-assisted, pure-currency hedge fund that delivered an annualized return of 14.5 percent and achieved a Sharpe ratio of 1.5. Just like other currency funds, however, the fund founded and managed by Thomas Jacobson (pictured) has been a victim of calmer currency markets in recent years.
Minastir AI Currency Fund is a directional hedge fund that takes long and short positions in highly liquid currencies. Jacobson added artificial intelligence to otherwise static systematic approaches of investing in foreign exchange markets. The fund’s trading approach is “fully automated and managed by our highly sophisticated AI trading algorithms,” Jacobson tells HedgeNordic.
In the simplest form, Minastir AI Currency Fund engages in intra-day and intra-week technical currency trading, employing traditional types of strategies such as trend, mean-reversion, breakout, or other price pattern strategies. Then “we also apply another layer to these traditional strategies, and that is where AI comes into the picture.” As Jacobson explains, “we take these traditional strategies and supercharge them using AI.”
“We take these traditional strategies [trend, mean-reversion, breakout, or other price pattern strategies] and supercharge them using AI.”
Jacobson reckons that “AI by itself is not a black box where data goes in and then magic happens, with strategies coming out.” The architect of Minastir’s approach argues that AI operates more like an assisting tool. “Our AI platform is a computer software with several machine learning models that we design and program ourselves,” he says. “First, we need to fundamentally define what instruments should be traded and how trades should be executed. The setup is then launched on our AI platform and the model is trained using our machine learning techniques on our high-performance computing cluster. Through the training process, the model learns and identifies market patterns within our pre-set frontiers.”
“AI by itself is not a black box where data goes in and then magic happens, with strategies coming out.”
AI enables static algorithms to operate with greater flexibility, allowing them to adapt to different market conditions. “Machine learning-based artificial intelligence allows traditional models with static parameters to use a broader scope of parameters, periods, and scenarios,” explains Jacobson. “In simple terms, AI is a method of optimization.” In addition to the ability to analyze massive amounts of data in a relatively short period, “different types of machine learning models discover different types of patterns and relationships in data that otherwise would be difficult or impossible for a human to detect.”
Risks Associated with Minastir’s Approach
“The biggest risk from our point of view is the external market events that are not scheduled,” acknowledges Jacobson. A politician talking about monetary policy, Donald Trump tweeting high market impact messages, or a key person from a central bank giving a speech at a university can affect currency markets quite significantly in unexpected ways. “From our point of view, this is the biggest risk because our systems focus purely on the technical currency trading,” explains Jacobson. “An external, fundamental impact changes the initial technical forecast the systems made previously.”
“The biggest risk from our point of view is the external market events that are not scheduled.”
Although Minastir AI Currency Fund’s investing approach only relies on price data, namely currency exchange rates, the fund’s systems are designed and trained to use major macroeconomic events “as a form of filter to reduce the risks when these events happen.” The fund reduces its trading activity around interest-rate announcements, monetary policy statements, and other scheduled events affecting currency markets. As the timing of these significant events and announcements are known well in advance, “our systems are able to reduce the risk ahead of these events, when trading liquidity is greatly reduced as well.”
With all the decision-making done automatically by Minastir’s AI algorithms, Thomas Jacobson and his team spend most of their time monitoring the existing models and developing new ones. “We monitor our models all the time and do check constantly if everything is working as it should,” explains Jacobson. “But we spend most of our time actually developing new machine learning models and seeing how combining different types of models and strategies can enhance performance.”
Past Performance, Future Expectations, and Competitive Advantages
Minastir Currency Fund delivered an annualized return of 14.5 percent since launching in October of 2012, but calmer markets have hurt the volatility-loving fund in recent years. The performance of the fund’s EUR share class, the share class launched in October 2015 and included in the Nordic Hedge Index, reflects the difficulty to generate returns in calm currency markets. “The performance in our first four years was particularly good because the markets were slightly different,” explains Jacobson. “Intra-day volatility was much higher, which was very suitable for breakout models.” In recent years, however, intra-day volatility in currency markets has dried up, and breakout models have suffered.
Yet, Jacobson reckons that moving forward, “currency markets will eventually return to their previous state. We see this coincidentally with the trade war blossoming up again.” Therefore, “we believe that the fund’s performance will increase as well to our historical averages.” Even if volatility does not come back to currency markets, Minastir AI Currency Fund’s algorithms are trained to adjust. “If the performance of breakout models decreases, for instance, the other models can compensate for that,” says Jacobson. “The system takes care of rebalancing itself.”
“Those who know AI very well might not have a very strong background within financial markets. And those who know financial markets very well might not understand AI that well. We have very strong knowledge both within AI and finance.”
AI represents a competitive advantage for Minastir compared to non-AI-assisted currency funds. As Jacobson explains, “AI works for us, but for others, it might not work so well. Others may not find how AI can add value to them.” Jacobson reckons that few understand AI and financial markets equally well. “Those who know AI very well might not have a very strong background within financial markets. And those who know financial markets very well might not understand AI that well,” argues Jacobson. “We have very strong knowledge both within AI and finance.”