I am a data-driven investor, a Carnegie Mellon alumnus, a Cornell graduate, and an engineer.

I’m not somebody that takes prevailing investing wisdom and assumes it’s correct. I’ve never been one to blindly trust axioms, such as “stocks with a low PE ratio are good" (a fact that I’ve later refuted with strong evidence).

However, when I analyzed the correlation between a stock’s fundamentals and their returns that year, I outright did not believe it. I double and tripled-checked the queries, broke them down piece-by-piece, and even used the most powerful language model to exist, OpenAI’s o1-preview, to confirm the accuracy. It’s right.

And what I discovered is shocking. For the vast majority of stocks, fundamentals do not matter.

What are stock fundamentals?

Fundamentals are a measure of how good or bad a company is. In finance, they can mean a lot of things, but for the purposes of this article, fundamentals will refer to earnings data, such as revenue, net income, and free cash flow.

These metrics matter because they tell the story of how fundamentally strong a business is. Unlike other asset classes (looking at you cryptocurrency), stocks are backed by the financial health of the company behind them.

Traditionally, we always thought this: companies that were strong fundamentally had strong stock prices later that year. We thought it was a 1-to-1 correlation – that if a stock increased their revenue or their cash flow, their stock price would increase at the same rate.

We were wrong.

What is my evidence that they do not matter?

Disclaimer: I am not infallible

I have strong evidence that these metrics fundamentally do not matter. But I am not infallible.

This is the part of the article where I want you to debate me. I’m not a SQL expert or a data scientist; I’m a software engineer. And thus, I might be doing an improper analysis. If I am, call me out, and embarrass me on the internet.

With that being said, I’ve taken statistics and graduated from good schools. I’m generally intelligent, and I wouldn’t be making such a strong claim if I wasn’t reasonably confident in my analysis.

So here’s what I did.

How I performed the analysis

I’ve created a text to SQL AI model that’s able to interpret financial data. Here’s how it works.

First, the large language model determines what exactly we’re talking about. There are about 10 different language model prompts, and we need to determine which prompt is the most relevant.

Then, if we’re talking about financial analysis, we will forward the request to that prompt.

The Financial Analysis Prompt is responsible for generating a SQL query against the database. The data primarily comes from the bulk downloads feature of the SimFin API.

Specifically, I transformed and uploaded all of the data into a BigQuery database so that I can perform the analysis. Then, I use the Financial Analysis prompt to generate my SQL queries and execute it against the database. This is called function-calling.

Once the query is generated, I retrieve the results from the database. The results are formatted and summarized, and sent back to the user.

In this case, we are asking the model to measure the correlation between stock prices and the stock’s fundamentals. This includes metrics like:

• Revenue: How much money a company generated in total
• Net income: How much of that money was actual profit
• Free cash flow: How much cash the company generated
• Return on assets: How efficient a company is at using its assets to generate profit

To measure these correlations, we had a back-and-forth conversation with the language model. We asked it the following questions:

For stocks with a market cap below $5 billion, what fundamentals was most correlated with stock price in 2021

We did the same for stocks with a market cap of $5 billion to $50 billion, $50 billion to $200 billion, and with a market cap above $200 billion. We also did the same for stocks in 2022 and 2023. We then analyzed the combined results to come up with our analysis.

Finally, we’re using the Claude 3.5 Haiku model for this analysis. This model is a great balance between cost and performance, and should be more than capable of helping us generate accurate SQL queries.

If you want to see the full conversation we had the model, check out the following link.

For example, here’s the query for stocks with a market cap below $5 billion for the year of 2022.

Take a few minutes and analyze this query. Send it to your data science friend, or maybe drag it into Anthropic or Open’s UI and ask it to analyze.

As far as I can tell, it is correct.

And, it tells a compelling story. Fundamentals do not matter.

What were the results?

If we read through the conversation across the years and across market cap, we come to one startling conclusion.

For most stocks, especially stocks with a market cap below $200 billion, the fundamentals do not matter. At all.

For example, for penny stocks, the correlation between the fundamental metrics and price is nearly 0. The highest correlation is 0.0184 and the low is 0.0010. For all intents and purposes, this is basically chance.

Mid-cap stocks had high variability, but generally have low to no correlations as well. For example, in 2021, EBITDA and Return on Assets had a correlation of 0.2 between price in 2021, but a correlation of less than 0.001 in 2023.

By and large, fundamental metrics had virtually no correlation between the stock’s price that year.

With the exception of one group of stocks.

Are there any exceptions to this rule?

Once the size of these stocks start to reach a certain threshold, we start noticing a certain tend.

These metrics begin to matter.

For example, for stocks with a market cap above $200 billion, the correlation between the stock price and metrics like net income, return on asset, and revenue is between 0.3 and 0.4 in 2022.

In 2023, it’s even higher, 0.43 and 0.8 for certain metrics.

These insights are extremely informative and tell a compelling story – fundamentals matter more the larger the stock is. Penny stocks and small caps seem to be driven by external factors, such as news sentiment or speculation.

In contrast, the mega-cap stocks are driven by how fundamentally strong the business is. These insights are unconventional and go against traditional investing wisdom, which is why I’m encouraging refutations, rebukes, and arguments in my comments.

What are the limitations to this study?

Before you start using a dart board and picking stocks to buy at random, I do want to say this – these results have significant limitations. Allow me to explain.

We did this analysis on an aggregate of stocks

For these stocks, we took the aggregate correlation by market cap. We didn’t do any additional analyses on other characteristics.

For example, for micro cap stocks, we may have found the biotechnology stocks had an extremely low correlation, but payment stocks had a moderately high correlation. Or, maybe stocks with a revenue below $1 billion had a low correlation and stocks with a revenue above $5 billion had a high correlation.

Another example – maybe this analysis had a low correlation for stocks that went up but a high correlation for stocks that went down. We wouldn’t know, unless we did the analysis.

Thus, while aggregate data provides a broad overview, analyzing subsets based on industry, revenue size, or other criteria can reveal more nuanced relationships, offering valuable insights for investors.

Correlation does not equal causation

In addition to the aggregate view of this data, an important fact with data analysis is that correlation does not equal causation.

I’m about to contradict myself just a little bit. Just because the data shows low correlations between fundamentals and stock price doesn’t necessarily mean fundamentals have zero impact.

Low correlation isn’t the same as no connection — it’s possible that fundamentals still play a role, but their effect gets overshadowed by things like market sentiment or external events.

This analysis is a starting point, not the full picture, so don’t ignore fundamentals entirely — they may still matter in ways that are just harder to see at first glance.

This analysis relies HEAVILY on SimFin data

This study relies heavily on SimFin data, and while it’s reliable, every data source has its limits. SimFin captures a broad range of financial metrics, but it’s not perfect, and some nuances might be missed.

This analysis is built on that foundation, so if the data has any gaps, they’ll affect the findings here. Keep that in mind before making decisions based on this alone — it’s a useful tool, but it’s best viewed as one part of a bigger picture.

Concluding Thoughts

When I invest, I don’t rely on traditional investing wisdom. I rely on cold-hard data.

In this case, the data told me a story that I thought I would never hear – for the vast majority of US stocks, the fundamentals have virtually no correlation between the stock’s price.

This effect fades as the stocks market caps get higher. But it is still present amongst the vast majority of US stocks.

As someone who religiously relied on stock fundamentals to make my decisions, this shocked me. I always thought that fundamentals mattered more than anything. But the data is telling me that I’m wrong.

Now, this study has significant limitations. The queries the model generated may be wrong or the data may be inaccurate. Some of the insights may be true today but false for 2025 and beyond. We should not rely on these insights for the sole basis of making financial decisions.

But at the very least, these insights are interesting. They go against the grain and force you to think about your assumptions before assuming its true.

“If a stock is below a $10 billion market cap, does its revenue increasing really matter?” you might ask yourself. “Maybe I should look more at the executive team and their experience instead of just the fundamentals”.

If I am wrong with my analyses or conclusions, I’m not one to intentionally spread misinformation. I want you to challenge me.

Post a comment and explain where I went wrong. Write on LinkedIn or Medium, and break down my arguments step-by-step. Create a YouTube video, and embarrass me on the national stage. Do what you have to do.

But as far as I can tell, my analyses seems sound. And you can replicate it for yourself. NexusTrade is a free, AI-Powered algorithmic trading platform that has these features implemented. You can do the analysis yourself with your own database and your source, or you can use NexusTrade and get started right away.

I think this demonstrates clearly the power of AI in the financial industry. Wouldn’t you agree?

Thank you for reading! By using NexusTrade, you can perform extremely detailed fundamental analysis using natural language. Want to try it out for yourself? Create a free account on NexusTrade today.

Follow me: LinkedIn | X (Twitter) | TikTok | Instagram | Newsletter

Listen to me: Spotify | Amazon Music | Apple Podcasts