A dedicated writer and digital evangelist.
Are you aware of how the buying and selling of stocks were carried out when there was no internet or computers? Back then, stock exchanges had active trading floors filled with brokers and traders. To make a trade or a purchase, they had to shout or use hand signals to alert others about their buy or sell orders. It looked a whole lot like an auction at a fish market today.
But then came computers and the internet to change the game completely. That not only replaced the archaic trading system on the floors of exchanges but also made it possible for individual investors to participate in markets on an equal footing. But that wasn't the last market-altering shift that would happen.
Next came the algorithmic or automated trading era - where computers are programmed to trade independently when specific criteria are met. This made it possible to execute complex trading strategies with pinpoint precision. Investors and brokers gained the ability to plot their strategies many moves in advance. And they could also protect themselves from downside risk with automated stop-loss orders.
The new trading systems also made a variety of useful algorithmic trading tools possible. They took automated trading to the next level by allowing computers to trade based on a variety of price signals analyzed in real-time. And since the computers can execute a near-limitless number of trades in an instant, high-frequency trading to exploit every minute price movement became the norm.
Algorithmic trading strategies also came with another advantage. Traders could use software like MetaTrader 5 and platforms like TradingView to backtest their strategies against historical market data. That meant they would know their strategy's odds of success before committing real money to it. It marked the final shift from intuition-based trading to data-backed analytical trading.
But algorithmic trading wasn't perfect. Using algorithmic trading, we can tell a computer to do A when B occurs. For example, an order to buy a stock (A) when RSI<30 (B). When condition B (RSI<30) is met, the computer performs action A (buys the stock). That's automated trading. Humans program and specify conditions in advance and the computer does everything else. But preprogrammed strategies can fall apart when confronted with contradictory or unexpected data. It's a vulnerability that has already spurred numerous mass-selloffs in the recent past.
Artificial intelligence (AI), however, goes a step further. AI trading solutions can train on multiple types of data to understand how and why markets move as they do. Applying AI to the task of stock trading can allow computers to figure out those A and B conditions themselves and get results with minimum risk. And they should also have the ability to shift strategies on the fly with no further human intervention.
The great thing about AI systems is that they can process gigantic amounts of information in mere seconds, far exceeding the ability of human traders. The fintech app Front, for example, built proprietary AI-driven algorithms to instantly analyze information from a variety of data sources. It considers historical company and stock data, international markets, news announcements and social sentiment, financial reports, and even stocks' compatibility with investors' existing portfolios.
The result is to quickly show investors the potential risk for any investment, so they can make more informed decisions in a fraction of the time. AI systems consider all the technical, fundamental, sentimental, and lots of other micro and macro factors to make the correct evaluation in real-time. They bring some of the art of the old-school intuition-based trader together with the power of data-backed analysis.
And stock trading AI systems are improving fast. This past June, researchers at the University of Cagliari reported the results of a project that used a convolutional neural network (CNN) to power a buy and hold stock trading strategy. The CNN trained on historical S&P 500 data and could predict the live market with 50% accuracy. And with some minor modifications, it did better than break even.
While it's still too early to declare AI the next big thing in stock trading, all of the early signs are pointing in that direction.
Already, basic AI systems can do a better-than-average job at predicting market movements. And they don't suffer from the tendency to trade based on emotions, as human traders do. They approach every situation with cool rationality – which means they can exploit market movements that run counter to the underlying data.
AI can also protect traders from fat-finger trading mistakes and work to stabilize markets, unlike algorithmic trading. A wide range of tasks (automation spectrum) can be performed without human intervention through AI that was impossible before.
With AI, Robo-advisors can evaluate millions of data points and carry out trades at the most favorable price points, and analysts can make market predictions with increased accuracy. And AI can also help large trading firms to mitigate risk while aiming at the highest possible gains.
However, to perform such functions, a large amount of data is required. An AI model can only provide superior predictions when the primary dataset it learns from is large and diverse. And developers must ensure to utilize training methods that don't leave the AI with blind spots or inherent biases. Only then can an AI-based trading system have a solid foundation upon which to build.
With that said, some factors could prevent AI from coming to dominate financial markets. As with any other type of technical innovation, there are bound to be missteps along the way. And considering the speed of today's markets and their growing dependence on automation, even a slight hiccup could cause havoc.
And there's also the possibility of AI being used for purposeful market manipulation. Back in 2010, a rogue trader used an algorithm to try and sway markets to their advantage – with disastrous results. With AI in the mix, such schemes might grow ever-more sophisticated, depending on what kind of fail-safe mechanisms evolve to constrain automated traders.
And there's still the matter of AI's immaturity as a technology. Even now, as impressive as AI trading systems are, they can't fully replicate human intuition. That could leave AI systems open to external manipulation.
For example, suppose an AI system tracks an influencer, Mr. X, using sentiment analysis. It would be up to its training data to inform a decision on how to weigh Mr. X's opinions. The AI can't yet factor in real-time contextual data.
However, a seasoned human trader might have known that Mr. X's followers, given his history, would discount his opinion and act accordingly – producing a contradictory result. And if an individual influencer carried enough weight with an AI, they might decide to try and influence that AI's decision-making for their own benefit.
The possibility of such outside manipulation – however remote – will remain a stumbling block for AI trading systems. That is until they evolve further and can make more intuitive decisions on their own.
What is clear, however, is that AI-powered trading holds several advantages over its algorithm-only predecessors. First, there's the fact that an algorithm-driven system is always limited by the skills of the person using it. Because they cannot conduct independent analysis, they won't always deliver profitable results.
An AI-powered system, by contrast, can make decisions based on rich data in real-time. AI can process massive amounts of data far faster than a human trader ever could, which increases the chances that it will make sound and reasoned decisions. And it can make those decisions fast enough to exploit even the tiniest market movements, or to counteract unexpected market swings by remaining flexible.
All of these factors mean that AI is already one of the best tools available to forecast which stocks to invest in. The risk-management benefits of current-generation AI tools alone make it more than worth using. And for that reason, it's no surprise that AI is already in use by the stock markets themselves for that purpose.
NASDAQ, for example, adopted an AI-powered anti-fraud system at the end of 2019. It's purpose-built to detect and defeat the exact type of market manipulation that caused the flash crash referenced earlier. And because the system can learn to spot other types of malicious activity, it may come to act as a check on some of the pitfalls of AI-powered trading identified here.
The bottom line is that AI has already revolutionized the stock market. It may not be as visible as the technologies it's displacing, but its influence is undeniable. And this is just the beginning. Moving forward, there's an excellent chance that AI will continue to evolve new feature sets to aid investors, brokers, and market regulators.
But even at this early stage, AI-powered trading tools can reduce the time, in-depth knowledge, and intuition necessary to build a strong portfolio for the long term. We haven't reached the point where human traders become irrelevant, but today's AI has closed the gap between casual investors and professional traders considerably. And given how fast that has happened – it won't be long before AI becomes the new foundation of the stock markets of the future.
Photos licensed via contributor's Adobe Stock account, by metamorworks, peshkov, AndSus.
Create your free account to unlock your custom reading experience.