While the financial world is still in the early stages of figuring out how artificial intelligence will work for — or against — it, one bank is putting the technology to work picking stocks for clients.
Openbank, the Spanish-based online unit of Santander Group
SAN,
SAN,
is offering its 2 million-plus customers across Spain, Germany, Netherlands and Portugal free access to a tool that uses AI to determine prices for Stoxx Europe 600
XX:SXXP
and the S&P 500
SPX
stocks.
The tool, which offers stock price targets for one, three, six and 12-month periods, has proven itself to be competitive over time, Gonzalo Pradas, director of Openbank Wealth, told MarketWatch in an interview on Wednesday.
Pradas said the bank began testing their AI tool’-s ability to pick stock prices in 2020.
He said the bank saw an average success rate of 30% for the models on which the AI project was tested at the time. Currently, the average success rate for the four term horizons that it is offering clients for stock picking — one, three, six and 12 months — is 56%, and Pradas said they got as high as 62.5% in one of those scenarios.
“This means that AI works, first, and second it optimizes the [success] rate that we’re used to, and this in four different term horizons, for more than 1000 stocks,” he said. MarketWatch does not have the details of exactly how they compiled those tests.
On its face, those percentages appear competitive with efforts of the mortals on Wall Street and elsewhere who were targeting other company performance measures.
A study published in the CFA Institute Journal Review in 2018 found that it can be tough for analysts to hit the mark. The study specifically gauged how often those EPS guesses would fall between minimum and maximum forecasts for a company within a specific period. The overall accuracy rate was 45% when the forecasts were broken into 100-day periods.
Openbank’s AI predictive models will be based on analysis of over 2,000 variables that are trained, optimized and updated daily based on past scenarios, helping to boost accuracy over time, the bank said. It has some experience here, launching Roboadvisor products for clients five years ago.
“We see and witness there are tools that facilitate the investing process or action. This is one more tool that we are adding and delivering to our clients with this objective,” added Pradas.
AI in charge of investing still seems a work in progress on Wall Street and in the financial world.
A report in The Wall Street Journal last August flagged 13 exchange-traded funds that put AI in charge of portfolios, with nearly all missing last year’s tech-led rally. The reason given for one fund’s underperformance was that AI had deemed stocks such as Meta Platforms
META,
too expensive.
Elsewhere, though, a study published last April by University of Florida researchers found that ChatGPT could predict stock price moves more precisely than some traditional models.
