Algorithmic trading with the use of Artificial Intelligence (AI) technology can definitely be of tremendous help at finding market signals, but can it replace human trading? In the following lines, we’ll try to understand whether machines are better at active trading than humans.
First of all, it’s worth mentioning that Wall Street heavily relies on algorithmic trading, especially in the stock markets. However, don’t forget an important aspect – the entire process of AI-based trading that involves big money is ultimately tracked by humans, who act as supervisors during the different stages and timeframes. While it is true that the human factor drives all the markets, many institutional investors prefer to implement some automated trading tools to mitigate the risks caused by emotions. Last week, Bloomberg reported that Ashok Krishnan, head of electronic trading at Bank of America’s global markets division, had collaborated with Phil Allison of Morgan Stanley and Mark Goodman of UBS to promote and develop machines focused on trading bonds and currency pairs rather than just stocks. This demonstrates that the trend is becoming prevalent even in markets that have been traditionally averse to algorithmic trading.
Some believe that all in all, humans win – if you don’t believe it, you can have your algorithm try beating Meir Barak the founder of tradenet in his live trading room. An experienced day trader and author of “The Market Whisperer” bestseller, he insists that active trading is not for robots. Barak inspires day traders via his YouTube trading channel, calling for active involvement in the trading process.
Electronic trading conquered the stock markets for real, but here is the thing – this concept might be valid for institutional investors who have long-term goals. Day traders and even swing traders should be somewhat skeptical of the idea that one can set some applications and go to sleep because the money would flow. The concept of passive income sounds appealing, but active trading requires a lot of involvement and even stress, especially in the case of beginners.
Let’s compare the day trading process with solving a problem – there is no predefined systematic approach to a problem. If it were a math problem with many restrictions, AI would be good at solving it based on predefined sets. However, the weakness of AI is that it wouldn’t be able to build the whole context of the problem – it won’t understand that the restrictions are restrictions.
AI might be ideal at automating several processes, but it cannot be aware of the context the same way it cannot be aware of itself – we’re far from the singularity, right? Humans are extremely good at understanding the context of different scenarios and how they can impact a given market or asset. This is because the context is not an intellectual problem but rather an emotional one, and, as we know, all markets are ultimately driven by emotions.
If you take your time and have a look around, you’ll easily note that we are constantly dealing with people in a world that is by definition spontaneous and unpredictable. Thus, the context of things is always changing.
Machine learning and AI are usually implemented for big data sets, which involve tremendous volumes of information. While this might help traders identify market signals based on programmed and predefined sets among large numbers of financial instruments, it is essential to understand that the correlation between these instruments is changing, and the AI wouldn’t always be accurate even at finding good entry points, not to mention active trading. Also, some AI-based models might work well during a certain market cycle, but they would eventually fail when the cycle ends.
The stock market is known for its volatility, internal variety, and spontaneity, which is why setting the AI to apply a set of features for two or more different stocks is quite difficult. No one denies the fact that strong correlations exists, but things become tricky when the context is not understood. The AI-based model would maybe explain and perfectly analyze the behavior of one stock while failing to do the same with the other stock.
This is actually what happened to the Black–Scholes model. The formula was implemented in the late 1960s by Fischer Black and Myron Scholes, who started an investment firm based on the model that they had developed. It worked well until market conditions changed and they lost a huge portion of investors’ funds. Note that this example related to the world of investing, but when it comes to day trading, things are even more difficult because of short-term volatility.
