Utilizing advanced algorithms and machine learning, AI tools like ChatGPT have opened the door to more nuanced and dynamic trading strategies. These strategies go beyond traditional data analysis, delving into sentiment analysis and predictive modeling, which have become increasingly valuable for today’s fast-paced markets. This evolution marks a shift from mere numerical analysis to a more holistic approach, where qualitative data, such as news headlines and market sentiment, have a pivotal role in influencing the development of trading strategies.
The trading strategy we will explore centers on analyzing news sentiment, a task ideally suited for AI tools like ChatGPT. The core idea is simple yet powerful: buy stocks of firms associated with positive news and, if not constrained, short-sell those linked to negative news. This approach leverages AI’s capacity to rapidly process and analyze extensive volumes of textual data, extracting sentiment trends that are otherwise difficult to quantify. By aligning investment decisions with the prevailing sentiment about a company, we can potentially capitalize on market movements driven by news.
- NAGA launches first trading app to be integrated into Telegram
- Zero commission trading is not the same as free
- Free ticket for the IX Trader Show in London on 27th September
However, this strategy is about more than just reacting to news sentiment. A fundamental component is maintaining market neutrality, which minimizes the strategy’s overall exposure to market movements. Market neutrality is achieved, if necessary, by balancing the strategy with a market index, ensuring that the focus remains on exploiting individual stock movements rather than broader market trends. Market neutrality is helpful, as it helps mitigate systemic risks and market volatility, making the strategy more resilient to market swings. Executing this type of strategy requires meticulous planning and nuanced comprehension of the financial markets and AI tools.
Profitable strategies have limited lifespans
We should understand that there are never guarantees when trading, and markets are increasingly becoming more efficient. What works during some months may stop working the next one since investors, especially active managers and hedge funds, are constantly adapting to the latest technology and looking to gain an edge in every trade. Therefore, adopting a dynamic mindset and understanding that any profitable strategy has a limited lifespan is imperative. Instead, the key is looking at the strategy not as a rigid blueprint but as a starting point for developing new investment ideas.
Coming up with successful investment ideas involves deeply understanding market microstructure—the specifics of how different securities markets work—and how the major players like hedge funds and market makers operate within the various rules and structures. The objective is to identify gaps where temporary opportunities are likely to arise due to the constraints, costs, risk exposures, or informational disadvantages facing certain common financial institutions. Large language models like ChatGPT can provide significant advantages in comprehending market microstructure by analyzing and explaining regulatory filings and exchange rules to identify participation constraints.
Analysing liquidity constraints
Liquidity refers to the ability to rapidly transact an asset or financial instrument without substantially impacting its price. Assets with high liquidity can be traded smoothly in sizeable amounts. Illiquid assets with lower liquidity are costlier to trade and have greater price frictions. Liquidity issues describe assets where trading large quantities is expensive. Analyzing constraints around liquidity provision can reveal areas prone to mispricing as institutions retreat. Small-cap stocks, over-the-counter bonds, distressed debt, exotic derivatives, and thinly traded ETFs are examples of illiquid securities.
Most institutional investors avoid trading significant quantities of illiquid securities because their large order size means finding enough buyers or sellers without adversely impacting the purchase or sale price is difficult. Since the existing resting buy and sell orders for thin securities lack depth, a big trade can drain the order book on one side, causing the price to spike or plummet before the total amount is executed.
This means an asset that initially looks attractively priced will rapidly become overpriced for the institutional trader as soon as they start buying it, eliminating their expected profit. Or, if they are trying to sell a position, the price will plunge too quickly rather than selling gradually at better levels. In essence, the very act of the institution trying to trade the illiquid asset makes the valuation deteriorate compared to the level they anticipated. This happens because their large order size quickly moves the market by draining buy or sell orders before crossing the total amount. For large investors, the potential market impact renders many compelling illiquid trades pointless since executing the full position predictably moves the price adversely before finishing. In contrast, investing a small amount of money is less likely to impact the price.
How Large Language Models can help traders
Informational asymmetries mean some investors possess better real-time information, data tools, or predictive models than others. This gives certain firms advantages in accurately valuing assets and risks compared to naive strategies that just extrapolate historical pricing. For example, hedge funds employing alternative datasets from satellites, sensors, online chatter, or web scrapers may have early signals of demand changes. Large language models can help us create and understand alternative datasets—text, images, and audio.
Specifically, LLMs can understand industry-specific data, extracting complex insights from unusual data such as public shipping logs, science publications, or satellite imagery that would otherwise require extensive expert programming. Their natural language abilities can also assess sentiment shifts around companies to predict revenue impacts of consumer attitudes or politics. LLMs can pinpoint relationships between information and asset prices by detecting patterns across massive volumes of leading indicators, economic data, news events, and more. This hugely expands the actionable alternative data for informational trading advantages.
Institutions limited by strict trading rules
Many institutional investors, like mutual funds, pensions, and endowments, have strict trading rules limiting which assets they can buy. For example, pension funds often cannot invest in risky derivatives or unlisted small-cap stocks, even if prices look attractive. This leaves gaps between the asset’s actual price and fair value in those niche markets.
Arbitrageurs (traders that detect and exploit price inefficiencies or discrepancies) would typically buy an underpriced asset, bidding up the price. However, complex assets attract fewer flows than usual, resulting in less smart money present. Specialized traders can take advantage of these fragmented corners that are ignored by constrained investors. Fewer players competing means pricing quirks and informational edges persist much longer than inefficient mainstream markets before being corrected. Investors can systematically exploit temporary valuation gaps and data advantages by focusing on assets excluded by institutional rule books.
Finding the best trading opportunities means deeply understanding how specific markets work and the limitations facing key investors. Where big institutions retreat due to illiquidity, regulations, or inadequate data—profitable gaps arise. Nimble traders focusing on these complex niche areas can leverage advanced alternative information ahead of constrained funds to exploit temporary investment opportunities. Fragmented corners enable pricing quirks to persist longer without sophisticated arbitrageurs closing gaps. By concentrating efforts on market segments defined by opacity, exclusions, illiquidity, and barriers to participation, agile investors can capture consistent informational and valuation advantages neglected by status quo players focused on only the most traded assets.
This is an edited extract from The Predictive Edge: Outsmart the Market using Generative AI and ChatGPT in Financial Forecasting by Alejandro Lopez-Lira (published by Wiley, June 2024)
About the author:
ALEJANDRO LOPEZ-LIRA, PHD, is an Assistant Professor of Finance at the University of Florida. His award-winning research is focused on machine learning and textual analysis and he has a lengthy background in finance from renowned institutions. He was the recipient of the 2023 BlackRock Best Paper Prize for the research underpinning this book.
