GPT-3 can do a lot of amazing things, from writing full-length articles to improving headlines, summarizing text, and much else. The system is great at finding patterns in complex, real-world data and making sense of them at scale.
For that reason, many GPT-3 users and prompt engineers likely wonder whether it can provide insight into one of the most complex systems of all: the stock market.
In short, can GPT-3 pick stocks?
Testing GPT-3 Stock Picking
To test this, I decided to have GPT-3 do two things:
- Look at historical data on the stock market, and explain why certain companies saw big gains
- Extrapolate from that, and predict which other companies would gain in the future
For data, I looked at the S&P 500’s list of top-gaining companies for the last six months. I chose a fairly long timeframe (rather than a single day) since a longer horizon should help GPT-3 understand any broader trends it can identify.
Daily movements in the stock market are often random–longer-term ones potentially capture more meaning.
Let’s take a moment here to remember that this is only an experiment. You should always consult with a professional advisor before making any financial decisions. This article is about testing an AI system–it’s not about providing actual financial advice you should act upon.
GPT-3 Stock Picking Prompt
Here is the exact prompt that I gave GPT-3 to train it for stock picking:
I reviewed a list of the top-gaining companies in the S&P 500 index, and wrote a financial analysis article about why those specific companies are gaining, and which companies are likely to gain value next. List of Top-Gaining Companies: 1. Ralph Lauren 2. Wynn Resorts 3. Halliburton 4. Schlumberger 5. Tapestry Article:
Again, the companies in my list are from the actual data on top gainers over the last six months. I used the davinci-003 model, which is the most up-to-date model for GPT-3 text generation.
Here’s the result:
It’s been an interesting year in the stock market, with many companies seeing dramatic gains and losses. One of the biggest gainers in the S&P 500 index this year has been Ralph Lauren, which has seen its stock price double in 2020. The company’s success is due to its focus on luxury fashion and its ability to navigate the pandemic.
Wynn Resorts has also seen its stock price double this year, thanks to its Las Vegas properties, which have seen a surge in demand since reopening in June. The company’s success is largely due to its ability to capitalize on the pent-up demand for leisure activities.
Halliburton and Schlumberger are two oil and gas companies that have seen their stock prices surge this year. The rise in oil prices has been a boon for both companies, as it has allowed them to capitalize on their expertise in the industry.
Tapestry, the parent company of Coach, Kate Spade, and Stuart Weitzman, has also seen its stock price double this year. The company’s success is largely due to its focus on luxury fashion and its ability to capitalize on the growing demand for online shopping.
Overall, these five companies have seen tremendous gains this year, and it is likely that they will continue to see gains in 2021. Other companies that are likely to see significant gains next year include tech giants such as Apple, Microsoft, and Amazon, as well as companies in the travel and leisure industry, such as Marriott and Delta Airlines. With the economy continuing to recover, these companies are likely to be some of the biggest gainers in 2021.
Results of GPT-3’s Stock Analysis
That’s a pretty cool result. Ignore, for a moment, that GPT-3 is making up the specific amount of the gains these companies have seen and the timeframe (“seen its stock price double in 2020”.) We didn’t give it any of that data, so it’s just guessing.
- Some of its analysis is spot-on. It says, for example, that Wynn has done well because of its Las Vegas properties. That’s actually true, as real financial analysis sites share that Wynn’s “US operations are going strong and in fact, delivering record results, the majority of their total revenue coming from the company’s Las Vegas Operations ($544.4M) branch as well as their Encore Boston Harbour ($211.8M).”
- GPT-3 correctly understands that the pandemic is behind much of this rise (and now knows the pandemic happened–initially it didn’t). It also correctly understood that high oil prices have buoyed Halliburton and other petro firms.
- GPT-3 says that Tapestry has done well due to demand for online shopping. That might or might not be true, but it’s a pretty good guess.
Overall, GPT-3 seems like it could easily get a job writing financial news!
GPT-3’s Stock Predictions
After it analyzes our real-world data, GPT-3 then turns to making stock market predictions for the future. Because GPT-3 writes in a narrative form, you have to do a bit of reading between the lines to get its actual predictions. But in essence, they are:
- Ralph Lauren, Wynn Resorts, Halliburton, Schlumberger, and Tapestry will keep performing well
- Apple, Microsoft, Amazon, and other tech companies will do well in the year ahead
- Travel industry stocks, including Delta Airlines and Marriott, will perform well in the year ahead
If GPT-3 was going to make a portfolio based on its analysis of the data, that portfolio would be:
How Good Are These Predictions?
Of course, it’s impossible to know if these are good picks or not until we see what happens with these companies over the next year! Some of them make sense–GPT-3 seems to assume that if a company is gaining quickly at the moment, it will keep doing so into the future.
That’s a reasonable assumption, if not necessarily an accurate one. In the real world, momentum investors pick stocks on this basis all the time.
GPT-3’s thoughts on the tech sector seem dubious since tech definitely appears to be in a bit of trouble at the moment. Who knows, though? Maybe the system is right, and a tech resurgence is on the horizon.
GPT-3’s predictions about travel stocks make a lot of sense, too. As the world reopens, more people may travel and stay at hotels. Of course, a recession would change that!
One thing I learned about this experiment right off the bat is that GPT-3 can write great financial news stories. Its analysis of the movements in the market seems at least as believable as the analyses provided by many real financial journalists.
How good is GPT-3 at picking stocks? Again, the system gave us a specific portfolio. Only time will tell how its picks perform!
To find out, you’ll want to subscribe to our newsletter. We’ll check back in on GPT-3’s stock picks in a few months to see how they did.
This article is for informational purposes only and should not be construed as investment, tax or legal advice. Always consult a professional advisor for advice specific to your situation before making any major financial decisions, and never invest more than you can afford to lose. Thomas Smith holds a diversified investment portfolio that includes a variety of securities and cryptocurrencies.