Select Page

Artificial Intelligence in Trading

Artificial intelligence in trading sounds like a logical next step for the world of finance.

And so far, we have seen many articles like this one. (Great read!)

One of the things we’ve come to appreciate the most in working with investors and traders that come from discretionary backgrounds is their excitement toward embracing new technologies. However, we often find ourselves playing “Mythbusters” when we explain exactly how those technologies are used. Artificial intelligence (AI) is no exception; everyone wants to use it, but not many people are aware of how it works.

Challenges of AI Investing

Artificial Intelligence (AI), within the context of trading, refers to simulating human-like learning, reasoning, and perception within existing system processes. Many people perceive AI as a robust, cross-disciplinary quality that leverages elements of computer science, psychology, linguistics, mathematics, and philosophy. Each discipline demands its own commitment of time, and resources, for a person to understand – and the same is true for an application. While a computer may have access to information, access is not the same as rational conceptualization. An AI still has to be taught to think, to reason.

Bringing an AI up to the same level of understanding as us is incredibly difficult, and in some disciplines can be impossible. This is not to say that it hasn’t been, at least, attempted. Fringe research groups, hedge funds, and asset managers have all tried to leverage the potential of AI, but to varying degrees of success. Firstly, determining the success of AI fund managers is difficult due to the fact that there are so few and the technology hasn’t existed for long enough. Secondly, success in financial markets is determined through an understanding and conformity with the collective truth. The market and its participants are under no obligation to behave in explainable ways, or with any deterministic fashion. Simply put, whatever the investors believe in will ultimately work. For many institutional traders, the benefit incurred from an AI does not outweigh its costs.

Artificial Intelligence in Trading: Applications

Although it may appear that AI within trading is that of legends, there are legitimate applications to AI within trading that are far less ambitious. AI can be a powerful tool in the right environment, especially one with fewer variables. In areas like market-making and risk management, AI is used to make time-sensitive decisions.

Most recently, one of the most widely used applications of AI has come from stealth trading and position management. In order to move large sums of capital in the markets, without arousing suspicion of front-runners and HFT firms, we can use AI to identify competitor trading activity, spot phantom liquidity, and run smart-ordering strategies.

It is better to have an AI that works well with a single task, than an AI that performs poorly across complex tasks.

AI for Daytraders

You might have come to the realization that there is little place for artificial intelligence for retail investors and day traders. Building a true cross-disciplinary AI isn’t realistic, but there are other ways to leverage the building blocks of artificial intelligence in augmenting your own decision-making processes.

In discussing AI and algorithmic trading we sometimes get lost in bringing order to all the moving pieces. An automated strategy is just that – automation of an idea that is discretionary in nature. AI aims to be proactive as well as reactive to other discretionary systems. The key to using AI as a day trader or retail investor is not in distancing yourself from the markets, but bringing the markets closer using technology. Here are some of the different use-cases we found with our clients:

  • Reducing slippage on order execution with smarter orders.
  • Combat spoofing and identify real volume and liquidity across order books.
  • Mapping retail trader volume and popularity.
  • “Fear vs Greed” indices backed by principles of behavioural finance.
  • Algorithm re-training.
  • Author identification and NLP-backed learning for EDGAR interpretation.
  • Risk conformant robo-advisory.

While there are significantly more applications to neural networks within trading algorithms, we find that these are the closest cases of significant artificial intelligence. Remember, algorithmic complexity is not a determinant factor of AI. In fact, most trading algorithms labelled “AI” lack any element of intelligence, but make up for it in robustness.

Read our blog post on how to build an automated trading system (trading algorithm).

How can I benefit from AI?

Introducing AI into your trading system can greatly benefit your performance, however, in most cases, it’s not necessary or possible. If you’re new to the world of algorithmic trading then you should start with the basics; adopting a systematic mindset, intelligence augmentation, and strategy automation. After all, the use of AI in conjunction with your trading system is only as effective as your strategy. Make sure you are absolutely certain of how AI will benefit your trading experience before you commit to its implementation.

Oh – and don’t worry. True artificial intelligence in trading is a long ways away.