Market Psychology often drives price movement in the currency markets. Two examples of this are greed at a price rally or fear at a price plunge, making speculators buy at the top and sell at the bottom. This kind of emotional behavior creates at times an inefficient market, with price patterns that repeat themselves over and over again, possible to exploit.
Historically, the use of statistical methods to exploit the price patterns has been successful for many hedge funds, mainly because these inefficiencies have been significant and easy to monetize on. When observing the same hedge funds in a more recent period, it is clear that their performance returns have decreased significantly.
There are likely numerous reasons as to why the hedge funds general performance decline has occurred. What we do know, is that the computational power, execution speed and access to information have dramatically improved amongst all market participants, and the overall level of competition is now higher then what is has been historically. The conclusion, after adding all known facts together, is that the inefficiencies that historically have been easy to monetize on now have disappeared, or at least become significantly harder to exploit.
"The Orchestra" - Our AI system
We believe that today’s market inefficiencies demand a trading system with multi-dimensional and flexible techniques to achieve stable return on investment. With our trading system, which is built using machine learning, we can achieve a deeper understanding of how the trading opportunities change and behave in different market dynamics. Furthermore, with a trading system that is built on a technological infrastructure, we can through the use of automated execution capture fast opportunities that would otherwise have been impossible to profit from.
Technology, specifically machine learning, is applied in CenturyOne in order to take optimized decisions and improve the decision making over-time. An illustrative way of describing how the technology works is to think of it as an orchestra. An orchestra constitutes of musicians that have trained to become experts on their respective instruments. But to achieve music that is more than the sum of its parts, a conductor is needed to steer and lead the orchestra as a whole.
The same circumstances apply to our trading systems. We have built an orchestra of agents, where many of the agents use machine learning to excel and become experts on specific tasks. The agents report to a supreme agent, the conductor, who makes optimized decisions based on the high-quality information provided by the members of the orchestra. The conductor is trained using a specific niche within machine learning known as reinforcement learning.
This creates a system with two levels of information, where the conductor only looks at the below level, the agents, and not the underlying data in itself. We have found this to make the conductor’s decisions more robust and not as easily affected by random events in the market.