1. The general impact of Artificial intelligence and machine learning on trade
Artificial Intelligence (AI) allows robots to take the place of people. In the 1980s, expert systems and fuzzy logic were at the forefront of AI research. With the decreasing cost of processing power, utilizing machines to tackle large-scale optimization issues became economically possible. As a result of advancements in hardware and software, artificial intelligence (AI) now focuses on the use of neural networks and other learning methods for identifying and analyzing predictors, also known as features or factors, that have economic value and can be used with classifiers to develop profitable models. This type of AI application is frequently referred to as Machine Learning (ML).
The use of methodologies for building trading strategies based on AI, both in short-term time frames and for longer-term investment, is becoming more popular, and a few hedge funds are particularly active in this subject. However, widespread adoption of this new technology is sluggish owing to a variety of variables, the most important of which is that AI necessitates investment in new tools and human expertise. Fundamental analysis is used by the majority of funds since it is what managers study in their MBA degrees. There aren't many hedge funds that depend purely on artificial intelligence. Although the use of AI in retail is increasing, the majority of traders continue to rely on methods introduced in the mid-twentieth century, such as classic technical analysis, since they are simple to understand and use.
It is important to note that AI and ML are employed in areas other than trading strategy development, such as building liquidity finding algos, and recommending portfolios to customers. As a result, as AI applications gain traction, the number of people participating in trading and investing choices diminishes, which has an impact on markets and price movement. It is too early to speculate on the overall implications of this new technology on the business, but widespread usage of AI may result in more efficient markets with lower volatility for lengthy periods, followed by periodic volatility spikes owing to regime changes. This is achievable because the influence of human subjective information interpretation will be reduced, as would the related noise. However, this needs to be shown in practice.
2. The Impact of AI and Machine Learning on Alpha Generation
Those that understand AI technology and the know-how to handle its dangers will have an opportunity throughout the early stages of its adoption. One issue with AI-based trading methods is that they might provide models that are worse than random. I'll attempt to clarify what I mean: conventional technical analysis is a losing trading strategy because techniques based on chart patterns and indicators get their rewards from a distribution with a zero mean before any transaction expenses. Some traders are usually located towards the right tail of the distribution, giving the mistaken appearance that these strategies are economically valuable. Longer-term profitability is difficult to attain in the futures and FX markets, regardless of approach, because these markets are intended to reward market makers. However, because of chance, some traders may make big profits in leveraged markets in a short period. Then, rather than luck, these traders attribute their success to their strategy and talents.
There are additional impacts with AI and ML, such as the bias-variance trade-off. Data-mining bias can result in methods that are over-fitted to previous data but fail on new data or techniques that are overly simplistic and fail to catch crucial signals in the data that have economic worth. This trade-off results in worse-than-random tactics and a negative skew in the distribution of returns of these traders even before transaction costs are included. In the post-quantitative easing period, this creates a profit potential for major funds and investors. However, when the worst-than-random AI traders are pushed out of the market and only those with solid models remain, the competition for profits will heat up. It is too early to predict whether AI traders or huge investors would prevail in this struggle.
I'd also like to address a frequent misunderstanding in this field: some individuals feel that the value is in the ML algorithms utilized. This is not correct. The actual value lies in the predictors utilized, which are often referred to as characteristics or factors. ML algorithms cannot detect gold in the absence of gold. One issue is that most ML specialists utilize the same predictors and attempt to create models in an iterative approach to achieve the best outcomes. This technique is hampered by data-mining bias and ultimately fails. In a word, data-mining bias arises from the risky practice of reusing data with many models until the outcomes in the training and testing samples are satisfactory. According to my study in this field, if a basic classifier, such as Binary Logistic Regression, does not operate effectively with a given set of predictors, there is a strong likelihood that there is no economic value. As a result, success is dependent on feature engineering, which is both a science and an art that involves knowledge, experience, and inspiration to come up with features that have economic worth, and only a tiny fraction of experts are capable of doing so.
3. Artificial intelligence and machine learning's impact on technical analysis
We must distinguish between conventional and quantitative technical analysis since all approaches that rely on price and volume series analysis come under this category. Traditional technical analysis, such as chart patterns, basic indicators, price action theories, and so on, was ineffective, to begin with. Except for a few imperfect initiatives of limited extent and reach, publications touting these approaches seldom disclosed their longer-term statistical expectations, instead offering simply assurances that if this or that rule is followed, there will be profit potential. Because market earnings and losses follow a statistical distribution, there have always been some who attribute their good fortune to these approaches. Simultaneously, a whole business sprung up around these procedures since they were simple to master. Unfortunately, many people believed that they could earn by being better at utilizing tactics that everyone else was using, and the consequence was a tremendous wealth shift from these naive traders to market makers and other well-informed professionals.
Some market specialists discovered in the early 1990s that a big number of retail traders were trading using these unsophisticated tactics. Some built algos and AI expert systems to predict formations and then trade against them, producing volatility that retail traders, sometimes known as weak hands, could not handle. In a broader sense, the failure of traditional technical analysis may be traced back to the elimination of strong serial correlation from markets beginning in the 1990s. It was primarily the strong serial correlation that gave the false impression that these procedures were effective. With a few exceptions, markets are currently mean-reverting, making simple technical analysis tools ineffective. However, some quantitative technical analysis tools, such as mean-reversion and statistical arbitrage models, as well as ML algorithms that exploit characteristics of economic value, frequently function effectively.
Because of the wide range of models and the fact that most are kept confidential, this sort of arbitrage is unlikely to be duplicated in the case of AI and ML, but the fundamental concern with this new technology is not confirmation bias, as in conventional technical analysis, but data-mining bias.
Observing the market and looking at charts, in my opinion, is becoming outdated. The future of trading is about real-time information processing, model development, and validation. The future hedge fund will not rely on chart analysis. Some people continue to do so because they are at a crossroads where old ways collide with new ones. Many traders who are unfamiliar with AI will struggle to compete in the future and will withdraw.
4. The new trading technology's winners and losers
The use of artificial intelligence (AI) will revolutionize trade in a variety of ways, and this is already happening. Investors may soon discover that once the present trend created by QE ends, medium-term returns will be far lower than expected. If this situation occurs, investors will be forced to resort to the traditional method of locating a skilled financial adviser who can recommend a portfolio mix and select stocks that will grow in value. In certain circumstances, the adviser will be an artificial intelligence program, and the procedure will be carried out online.
Traders must become acquainted with this new technology. Most traders are still battling with outdated approaches, hoping that "buy the dip" would continue to work and deliver profits for a few more years.
One of the issues is the moral hazard fostered by the central bank with direct financial market assistance during the previous eight years. Many traders and investors now feel that bear markets are impossible because the central bank will be there to transfer their losses to everyone else while keeping their gains. As a result, the majority of market players are unprepared for the next major market regime shift and may suffer catastrophic losses.
There are several online resources for learning about machine learning, artificial intelligence, and trading. The greatest approach to learning is to tackle a few practical situations. However, I believe that most traders will be unable to adjust. The combination of abilities necessary to comprehend and deploy AI eliminates 95 percent of traders who are used to drawing lines on charts and watching moving averages.
Investors should conduct their research and seek the advice of a professional financial adviser who is conversant with these new developments. Every investor has a unique risk aversion profile, making it impossible to provide broad suggestions. There will be a profusion of Robo-advisors soon, and choosing one that meets specific requirements and objectives may prove difficult.
Anyone who is unfamiliar with ML and AI and their applications in trading and investing may find it more beneficial to contact a professional who is knowledgeable in this field than beginning on a path of reading books and articles, which may be done once the fundamentals are grasped.
Comments