Jan 18, 2023

8:00 AM MYT The age of the human trader could be fading now that artificial intelligence is playing an increasing role in helping investors identify market trends and determine what stocks to buy and sell. — Photo by Nicholas Cappello on Unsplash A handful of money managers are banking on a robot that works continuously day and night, but only gives its advice on stocks once a week - each Monday morning before the opening bell on Wall Street. The algorithm – built by the North Side tech firm StockSnips – is trained to read and process more than 50,000 media articles a day before picking stocks based solely on how companies are being talked about in the news. Chris Cannon, of Bridgeville, is one of 14 financial advisors and hedge fund managers testing the artificial intelligence in its early stages - using mostly their own money, although some have handpicked a few clients willing to take the risk. Each Monday morning, they anticipate receiving a trading file from StockSnips. They only need to trust the machine and follow its instructions to the letter. "We have control of the assets. They just send us the information to execute the trades," said Mr Cannon, owner of RetireRight Financial Planners in Bridgeville. He has made weekly trades along with the algorithm for six months. "We like what we see," he said, "As it continues to perform, we'll continue to allocate more in that direction." AI and stock picking The age of the human trader could be fading now that artificial intelligence is playing an increasing role in helping investors identify market trends and determine what stocks to buy and sell. Robots have advanced to the point where they can predict stock movements with a high degree of accuracy given their ability to analyse massive amounts of data and spot patterns that would be impossible for humans to decipher. Major Wall Street players like Black Rock have invested heavily in in-house artificial intelligence. Other firms are following its lead. According to a 2020 report by the US Securities and Exchange Commission, 78% of stock market trades were performed by trading centers that depend on automated centers and algorithms. The algorithms are programmed to perform functions that constantly search for the right trading conditions - such as oversold stocks or rallies. Bots can execute trades to buy and sell automatically. StockSnips has sought to gain an edge in the stock market by programming a machine capable of converting news articles into an actionable sentiment signal, a score based on how a stock is perceived in the media. The program reads everything published by the Wall Street Journal, Barron's, Benzinga, the Associated Press, analysts' reports and Securities and Exchange Commission filings. Then it generates a sentiment score for each individual US stock. Social media references are not included. The algorithm is programmed so that the most highly ranked stocks are bought and those stocks with the lowest sentiment scores are sold. "Markets are driven by sentiment," said Ravi Koka, CEO of StockSnips. "It's not just whether the company met its numbers, or revenue and earnings. But it's also investor emotions that drive the prices. "If you saw what happened to Tesla or GameStop, you saw the valuation is driven by the company's reputation and the sentiment of the investors." Up against the markets The volatile stock market last year produced the worst losses for investors in more than a decade. StockSnips outperformed the S&P 500 index by losing substantially less money. While the S&P 500 lost 18.17% in 2022, StockSnips held its portfolio losses to 5.6%. The 3-year annualised return for the S&P 500 at the end of December was 7.72% compared to 17.38% for StockSnips. The model had only ever been tested in a bull market since 2016, but in 2021 the software determined on its own that the time had come to shift sectors. "At the end of 2021, we saw that the models were selecting stocks that were shifting out of tech and growth sectors, and moving into particular stocks that were more in line with consumer staples, energy and some financial," Mr Koka said. A data scientist and serial entrepreneur, he took SEEC Inc, a company he founded, public in 1998. At its peak SEEC Inc had about 200 employees working in offices off Montour Run Road near Pittsburgh International Airport. The company moved its headquarters to New Jersey after being sold. Mr Koka said StockSnips is so advanced that its history of knowledge is maintained and built on so the sentiment score is a composite of past and current sentiment. "Sentiment has memory," Mr. Koka said. "You still remember what happened last week. But today's stories will further influence you either positive or negative. "An extreme example of that is when Boeing had the big air crash and the stock tanked," he said. "That didn't go away overnight even though Boeing put out a lot of stories about how they're addressing the aftermath." Boeing stock dropped after planes it manufactured crashed in Indonesia in 2018 and China in 2022. Mr Cannon manages about US$200mil (RM865.70mil) in client assets. He said the small number of his clients with money allocated to StockSnips are fully aware that a computer is at the steering wheel. He believes the automated system is worth introducing to more clients and playing a bigger role at his 20-year-old firm. "Most of our clients – I'd say 75% – are approaching or in retirement," Mr. Cannon said. "The other 25% are savers, working to get to retirement. "My focus is not so much on upside capture, but minimising drawdown and allocating to the appropriate areas that are not getting hit the hardest." Last year, marked a significant change in trends for the stock market from a growth market to a value market, and StockSnips shifted gears when the time was right, he said. "My job as a planner is to put my clients in the best place to succeed, and I do believe this technology has some application to do that," he said. Converting stories As someone who worked with StockSnips to develop the software, Frank Li was the first financial adviser to use it when it launched two years ago. His investment strategy is more similar to billionaire Warren Buffett's long-term buy and hold method. Fundamentals don't change weekly or monthly, but he believes sentiment does drive the market in the short term. "We have tested it the last couple of years, and the performance has been really, really good, especially on the downside protection," said Mr Li, owner of Kailasa Capital Management in the Strip District. He said he has about 20% of his personal wealth riding on the StockSnips strategy. He is being far more conservative with client money. He estimates that 2% of the US$450mil (RM1.9bil) assets under management at his firm is with StockSnips. "We feel comfortable starting to introduce our clients to this strategy," he said. He starts by selecting 220 companies from the Russell 1000 companies. Then using the StockSnips signal information, he ranks the 220 companies and the top 30 will be selected on a weekly basis. "That's how we manage every week when we manage those 30 stocks," Mr. Li said. "We make more trades, but not necessarily take higher risks." Currently, StockSnips only licenses its model through registered investment advisers. But the company does have a tool for individual investors in the form of a mobile app that will synthesise what is being said about a stock. Essentially, the app provides users with all news articles, positive and negative, about any particular company that the user is interested in researching. The professional model provides the sentiment scores assigned to those stocks that the algorithm predicts will be winners in the next week or month. "Our algorithms convert a story. The stories are about stocks," Mr Koka said. "We take that information and convert it into an actionable score. That's what we are leveraging to navigate the markets. "The other thing we've proved is sentiment is a predictor of price movement," he said. "It's a leading indicator, not a lagging indicator." – Pittsburgh Post-Gazette/Tribune News Service Article type: free