AI has made it much cheaper to investigate the market hunches I’d normally leave sitting in the back of my head.

You notice something odd and wonder: has this happened before, and does it actually mean anything?

This week, Meta gave me one of those questions.

It reported earnings a week earlier than its usual pattern. Then the stock fell sharply.

Was the company trying to rip the Band-Aid off? Do companies sometimes report early to get bad news out of the way?

Of course, you could tell the opposite story. Maybe a company reports early because it’s eager to share good news. Or maybe we’re giving the calendar too much credit.

I wanted to know whether there was a pattern beyond Meta.

So I used AI to help test roughly ten years of S&P 500 earnings reports.

Here’s how I framed the research and what I found.

​Download the full report here to follow along.

Was this email forwarded?
Subscribe to get more AI investing breakdowns just like this!

Methodology

First, we needed a definition of “early.”

We estimated each company’s usual reporting date from its own earlier releases. A report counted as a week early if it arrived at least five trading days before that date.

That’s different from changing an already announced date. We were looking for departures from a company’s normal reporting pattern.

Then I wanted to check two things: did the company beat consensus EPS, and did its stock outperform the market when the results came out?

​

For the stock comparison, we measured the reaction-day return minus SPY’s return. A stock going up isn’t especially impressive if the market went up more.

We compared early and late reports with on-time reports from the same reporting seasons.

We used the companies that belonged to the S&P 500 at the time. Meta’s latest release stayed out of the test because it prompted the question.

If you want to try out the prompt for yourself: earnings-date-research-prompt.pdf​

Results

1. Early reporters beat EPS less often in this sample

Companies reporting a week or more early beat consensus EPS 72.7% of the time, compared with 76.0% for on-time reports.

Very small correlation for companies moving up their earnings tending to miss. But difference and statistical significance is small enough where it's inconclusive.

​
2. Their stocks did better on average

Early reporters averaged a +0.52% reaction-day return relative to SPY, versus −0.09% for on-time reports.

So the earnings result and the stock result pointed in different directions. Fewer EPS beats, but a better average share-price reaction.

The return difference was also inconclusive. The range of plausible outcomes included both underperformance and outperformance. I wouldn’t buy a stock just because its earnings date came early.
​

3. Late reporters were less likely to beat EPS

Companies reporting a week or more late beat estimates 68.0% of the time, versus 75.5% for comparable on-time reports.

The earnings association was clearer here. I’d be interested in looking at what else was happening at those companies when they reported late.

But the lower beat rate didn’t establish a stock-return edge. It wasn’t enough to justify a short trade either.

What I'd do with this

I’m leaving the earnings date out of the trading decision for now.

But the important takeaway is this:

If you have a market hunch you’ve never tested, try it with the prompt above. Give AI the context, have it plan the research, and work through its questions before it starts collecting data.

AI allows us to now test these market hunches at scale.
​
By the way, if you want to read the full report, check out the report here.

Personal

I'll be spending more time in Singapore in mid-November working with our corporate clients. Let me know if you're around in Singapore! Coffee's on me :)