AIconomist Hub

AI in Todays Investing Scene

Over the past year, artificial intelligence (AI) has made a strong appearance in content creation and the automatization of simple tasks. However, when it comes to investing, things have been pretty quiet. At least from my perspective, there hasn’t been much buzz about AI related visions in investing, especially not from the viewpoint of private investors.

In reality, AI has been used in investing for quite some time through different kinds of algorithms. Particularly, large funds have highly automated buying and selling functions to respond very quickly to market changes. These algorithms are often referred to as ‘quantitative’ or ‘algo’ trading systems. They collect a tremendous amount of data from the markets, allowing them to detect even small changes very efficiently.

The volume of data is so big that it’s beyond human capacity to perceive these changes independently. Some individual investors have also developed their own algorithms, but they are so complex that only a few can create an algorithm that’s reliable and high-quality, making them still quite rare.

In practice, these algorithms and AI, in general, operate by being fed huge amounts of historical market data, such as price and volume data, on which they try to predict future movements and potential buying opportunities. In the economy, almost everything is related to everything. For instance, a change in inflation usually leads to a change in interest rates, which affects borrowing, which in turn affects a company’s financial situation. And this is yet a very simplified example.

AI can accurately detect how a change in inflation has reflected on the general financial health of companies, and it can identify buying and selling opportunities before the rest of the market, thus giving its user an advantage.

AI is also efficient in risk management, helping to identify financial risks that are otherwise hard to detect. However, they cannot predict future risks that are completely unique in nature. For example, the impacts of the coronavirus pandemic could not be predicted because such a situation had not been encountered before in the light of current knowledge.

You must be careful with AI, as it should not be trusted on its own. A good amount of human thinking and observation brings a good and necessary balance to decision-making situations.

Algorithms and Machine Learning in Investing

Above, I briefly told you about the principles of artificial intelligence in investing, but let’s get a bit deeper into algorithms and machine learning. Let’s go through a few more advanced functions. Some algorithms are related to what is known as real-time market depth analysis. These algorithms identify market liquidity, that is, the number of buyers and sellers. Adequate liquidity is required to execute a large trade without affecting the price too much, and therefore it’s very important.

Algorithms thus examine order book data and its changes to predict future price movements. This creates opportunities to make trades at the best possible time. For example, an algorithm notices that the number of sell orders for a particular stock has significantly increased. This could signal an upcoming price drop, allowing the algorithm to sell the shares before the drop happens, protecting the portfolio from larger losses.

Another advanced feature of algorithm utilization is complex trading strategies. In this function, machine learning is used in portfolio construction so that it assembles a well-diversified portfolio, but in such a way that the investment targets correlate as little as possible with each other. This reduces the overall risk of the portfolio.

For example, if you invest yourself in many German companies for diversification purposes, they are still all subject to market risk because they are all located in the same country. The same can be thought to happen when diversifying among cryptocurrencies, or investing in companies within the same industry sector.

The algorithm aims to minimize this risk. At the same time, it creates advanced models for timing the markets based on recurring phenomena. As with all machine learning, the more data that accumulates, the more efficient the algorithm becomes in this. These models can detect the beginnings of trends in different sectors, then weighting the portfolio towards these rising trend markets and reducing weight from less promising sectors.

For private investors, there are now many different digital platforms, so-called robo-advisors, that they can utilize in their investing. Robo-advisors are like digital, algorithm-driven fund managers, essentially doing the same as human fund managers have done. However, a digital robo-advisor can serve many more customers and without a salary, so its costs are much lower than those of traditional fund managers.

Robo-advisors first create a profile of the investor through questions, based on which they create an investment plan that serves the client’s goals. They also balance portfolios according to market movements and are even capable of considering tax benefits!

A very good option for people who don’t have the time or interest in managing their own portfolio. I recommend finding out more about robo-advisors if you recognize yourself, especially if you are just starting out as an investor! However, if you want personal advice, or if your investment strategy is more complex, robo-advisors may not be suitable for you.

Risk Management

Earlier, we indirectly talked about risk management through diversification of portfolios and the anticipation of sales. However, let’s look more closely at the topic, as risk management is a big part of investing.

One very popular tool in risk management is Value at Risk (VaR), which allows an investor to assess potential losses in their portfolio at a certain confidence level. There are a few different methods for detailed calculation, but generally, the calculation of Value at Risk is based on defining the maximum potential loss at a certain confidence level (usually 95% or 99%). It tells the investor what the maximum monetary loss is at a certain probability.

If the confidence level is 99% and VaR gives a result of $10,000, it means there is a 1% chance that the loss would be bigger than that $10,000. At a confidence level of 95%, the probability of exceeding the amount would be 5%, etc. The assessment is also made for a certain period, such as a week, month, or year. The longer the period, the greater the risk.

An AI, having taken in a lot of financial data, thus has really good possibilities to calculate these probabilities and amounts much more efficiently than a human. With AI, the investor stays much better informed about the risk they are carrying. Here, as in everything else, it should be noted that the data is based on past information, so it cannot be blindly trusted, but it gives the investor very important information to support their own analysis and minimizes the risk that a human has overlooked some part of the data.

Risk management assisted by AI is thus largely based on extensive analysis of historical data and minimizing risks based on it. The role of humans always remains to estimate near future changes themselves, even though AI provides good guidelines for identifying trends in the near future, but it does not always take into account phenomena in the constantly changing world.

The Role of Data in Analyses

The comprehensiveness of data is critical for AI-based models, as the more accurate the analysis desired, the more data the models need. If analyses are made based only on the past five years, the results can be really distorted, as there is a very high possibility of factors distorting the data over such a short period.

For example, if we were to make decisions now based on the past five years, it would include the coronavirus pandemic, the Russia and Ukraine war, the Gaza war, and massive government stimulus packages. These are not normal situations when viewed over the long term, so data from these times should be approached with caution.

This data from these uncertain times itself is very valuable for detecting deviations, but it needs to be supported by a lot of data from ‘normal’ conditions to make real comparisons. The best decisions are made with the support of AI, but still for a long time through the strong expertise and critical thinking of humans.

Future insights

Artificial intelligence creates really strong opportunities for future innovations in investing as well. The current AI-based tools of large organizations are unlikely to change much due to this new AI hype, but rather they will continue on their own strong path.

I see the new tools bringing great opportunities for private investors to develop their own analyses and achieve better returns. The algorithms and models discussed in this text are not possible to obtain or copy for personal use, but they are intended to spark thoughts about possible uses of artificial intelligence. Data plays a big role, and private investors can now collect their own data banks much more easily with current tools than before.

An example of this is ChatGPT’s ability to read PDF files, which can be utilized to gather data more efficiently from companies’ materials. This material can then be tidied up with the help of Microsoft Copilot in Excel, which can again be fed to ChatGPT for interpretation, preferably with a finely tuned financial GPT for oneself. I plan to delve deeper into the tools and their utilization possibilities in the future, so be sure to read my posts also in the future if the topic has sparked your interest. Thank you for reading and see you soon!