The Economy Rocket

The Economy Rocket

Share this post

The Economy Rocket
The Economy Rocket
Can ChatGPT Pick Stocks Better Than My Algorithm?

Can ChatGPT Pick Stocks Better Than My Algorithm?

I Put It to the Test

The Economy Rocket's avatar
The Economy Rocket
Nov 08, 2024
∙ Paid
4

Share this post

The Economy Rocket
The Economy Rocket
Can ChatGPT Pick Stocks Better Than My Algorithm?
2
Share

A warm hello to all new readers! I’m glad you are on board.

New here? I’m Michael and the founder of TheValueVantage.com and on a mission to make beating the stock market as easy for you as choosing your favorite ice cream.
How it all started.

As technology races ahead, we’re left asking an age-old human question in a new way: can machines really outsmart human insight?

There’s only one way to find out ...

In the stock market, everyone’s looking for an edge. Some people go with their gut, others rely on complex formulas, or gurus, and lately more and more investors are turning to AI. I’ve spent over two years fine-tuning my stock-picking algorithm, and I couldn’t help myself and wonder: how would my algorithm stack up against one of the most advanced AI tools in the world?

So, I pitted my algorithm against ChatGPT, the AI that many folks have praised for everything from writing code to answering tricky questions. I wanted to see if it could tackle a task that’s not only technical but also notoriously unpredictable.

Could ChatGPT, with its processing superpowers, analyze stocks like a Chartered Financial Analyst (CFA) and make solid buy predictions? And if so, could it outperform my algorithm?

How I Set Up the Test

I provided ChatGPT with a list of 25 stocks chosen by my algorithm and the full set of quarterly fundamentals and price data from the 12 months before purchasing them.

The task for ChatGPT was to assess each company’s financial health and provide a buy, hold, or sell recommendation. I wanted it to approach the data the way a CFA might—looking at both the numbers and the risks to forecast where the stock might go over the next 12 months.

ChatGPT processed all 23 stocks in about 10 minutes, which is fast. Why not 25? It couldn’t analyze two of the companies because of “issues with the data”.

Numbers Don’t Lie

Here’s the meat:

  • ChatGPT correctly predicted the stock price direction 5 times out of 23, giving it a 21.7% success rate.

  • It was unsure 6 times, meaning it didn’t give a clear buy or sell recommendation.

  • It got 12 predictions wrong.

And here’s how my algorithm fared:

  • It correctly predicted the price direction 18 times out of 25, for a 72% success rate.

  • My algorithm’s picks generated a 45.4% portfolio return over the 12-month period, outperforming both the S&P 500 (which gained roughly 38.4%) and the Dow Jones (up 28%) over the same timeframe.

In short, TheValueVantage.com beat both ChatGPT and the broader market.

Side note: Three of my picks—CHK, CPE, and SLCA—were acquired by other companies. We would have profited from the merge. After these mergers, I marked them as “sells” because they essentially stopped existing in their original form.

Here are the stock price charts. Left of the green vertical line is the assessment window. To the right is how the stock price evolved post-purchase.

Why Did ChatGPT Struggle?

It’s important to understand the intended function. My algorithm is “custom-designed” to evaluate stocks. It’s fine-tuned to analyze fundamental data, spot patterns, and make the call—all based on methods that have worked well for way over 27 years. ChatGPT is a general-purpose AI. It’s mostly trained on language, not finance.

Six times, it gave an “unsure” response, where it didn’t recommend either buying or selling. ChatGPT’s reluctance to commit made it less effective as a stock-picking tool.

ChatGPT Has Potential, But It’s Not Ready for Wall Street

ChatGPT isn’t ready to replace human analysts or a dedicated stock-picking model. It has impressive reasoning abilities, but with only 5 correct picks out of 23, it’s not something I’d bet my money on.

That said, we all know AI is advancing quickly. The fact that ChatGPT can even attempt something is impressive, and it’s likely that future versions will only get better.

This experiment has given me some new ideas for improving my algorithm. AI isn’t standing still, and neither am I. Rest assure, my model will keep developing to stay ahead—and now that I’ve seen what ChatGPT can (and can’t) do, I’m happy to keep pushing forward.

For My Paid Subscribers

You can download the entire Excel sheet with all the 23 ChatGPT’s detailed analysis. Below my picture is a button that starts the download.

It’s interesting to read through—like most analysts, ChatGPT sounds very convincing, even when it’s more wrong than right.

Nuggets

I think we writers should support each other. That’s why I will give two shoot-outs this week.

  • 8,000+ investors start their week with Money Machine Newsletter's insights to get smarter about investing in stocks. It's free, it's fast, it's a no brainer - just your weekly dose of market beating stocks in a 5-minute read.

  • The Intellectual Investor contains many articles on a variety of non-investing topics, from classical music, to travel and parenting, to Stoic philosophy.

  • And a honest thank you to Kevin from Atmosinvest for his great feedback on my last article. Loved it. Give him a follow as well, as he is a great guy.

  • And last but not least. I’ve listened to the feedback and lowered the price for TheValueVantage.com. At least for the beginning. Check it out.

Take care,
Michael

Hit the heart if you believe in human power!


This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Michael Pavlovic
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share