Financial trading

There are quite a few quantitative trading companies that invested heavily in their Digital infrastructure, aiming to be the first movers in the highly sophisticated area of High Frequency Trading (HFT). On top of those names, is Domeyard. In the below article, we explore the challenges faced, the limitations imposed and the future of HFT in light of the latest Artificial Intelligence and Machine Learning applications. 

Like many Hedge Funds, Domeyard faced two broad types of challenges, as per the below:

• Operational challenges, such as:

1. Physical challenges: Early on, Domeyard struggled with arranging back up power and managing the heat and noise coming from the servers. This issue was resolved later on by moving the servers to Data Centers.
2. Legal challenges: Hedge funds in general are subject to many legalities and regulations. In Domeyard’s case, the challenge was in dealing with the legal types of entities supporting the funds as well as the tax implications. 
3. Technical challenges: Collecting and writing data was challenging due to the different and continuously changing systems and individual exchanges. In addition, with speed of processing being the most important factor in the success of HFT model, the slow-down of systems that results from combining execution servers with data servers is considered to be another challenge.

•  Strategic challenges, such as:

1. Fundraising: It is normal to face challenges in raising funds; however, with Domeyard, the challenge really was first to explain the sophisticated product to Tech investors, and second the inability to provide a prototype; something which most investors understand and trust. The value however, was proved later on, by the $5m investment made by a well-known fund manager.
2. Structure: As a hedge fund, an inherent challenge is the structure of fees, between the management company and the fund. These fees included Management fees and performance fees. In addition, Domeyard was expected to be relatively comfortable with it’s expected return and target sharpe ratio without a live track record.
3. Cost: Despite Domeyard’s expertise and competitive advantage in building their software internally; however, to cover their fixed cost they still needed to trade around 10,000 trade a day. In addition to the high trading costs, Domeyard’s purchase of CME equity membership costed them a fortune. Yet, through this purchase they are aiming to provide the investors with a higher level of comfort as well as build Domeyard an asset; a gift that keeps on giving.

According to the article, “Potential and Pitfalls of Artificial Intelligence in the Trading Environment”, there are several limitations of applying artificial intelligence and machine learning to financial trading, such as:

1. Inaccuracy of data: First of all, a major limitation in the world of AI and ML is “finding sufficient high-quality data to feed and train systems”.
2. Algorithmic behavioral bias: I find this factor to be very interesting, where people question the effectiveness of algorithms, if they were trained in periods of low volatility and used in periods of high volatility, and vice versa.
3. Lack of autonomy: It is noted that although AI provides the autonomy to choose where and when to buy or sell, it does not offer the autonomy to choose what to buy or sell. This puts more pressure on human intervention and requires more effort to run the platform and train the machines.

According to Christina Qi, co-founder of Domeyard, another limitation is:

1. Slower processing: “Deep Learning”, although very advanced and useful for unstructured data purposes, yet drives the system to become very slow. Therefore, leveraging on machine learning might do more harm than benefit, given the importance of speed in HFT model.

The importance of highlighting the limitations always lies in what’s next; Learning. In light of the ever changing nature of Artificial Intelligence (AI) and Machine Learning (ML), the future of HFT looks promising. Below are the major areas where AI and ML are expected to play a critical role in the success of Hedge Funds:

1. Autonomy of choice. As mentioned earlier, the lack of autonomy of choice is a limitation in the current system. It is expected that this would be permitted in the future, under Supervised Machine Learning (Potential and Pitfalls of Artificial Intelligence in the Trading Environment, 2018).
2. Foreign Exchange Rate Prediction. Another application would be forecasting FoRex rates, through Support Vector Machines (SVMs), a type of Supervised Machine Learning. According to the research paper published in 2018, “FoRex Trading Using Supervised Machine Learning”, “The net profit of using SVM in add-on robot is higher than the one without using SVM” and “Robotics using SVM (…) can prevent the capital reduction compared to the normal Robotics’ transactions”. I find this to be very helpful for the world of trading, as it could help traders make better decisions not only based on historical information, but more importantly based on predictions and insights. (Thu & Xuan, 2018)
3. Strategy Selection. As Christina Qi - Chairman of Domeyard- states, machine learning can be applied in strategy selection. This probably includes simulating different scenarios and using AI and ML to pick and choose the one with the best outcome. Taking this one step further, ML can help switch between strategies (active and passive) as and when the conditions are predicted to change, in favor of the investor. The more the machine learns, the better the outcome, and -maybe one day- the closer it’ll get to what each investor is looking for individually, according to their set of morals, interests, etc.. and not necessarily only according to their Financial goals. This would be a very interesting area to explore in the field of Behavioral Finance.

The time where efficiency was the number one targeted outcome of any investment is far gone. Predicting the future is what most investors would like to see, especially in the highly lucrative and equally risky field of High Frequency Trading. Having such smart systems would definitely support traders in making better decisions, and more importantly, it would pave the way for creating the best platforms to predict the performance in highly uncertain environments, like the ones we are witnessing today.