New ETF uses AI to help investors buy the dip

"Buying the dip" is a popular trading method, but can AI determine when there is a dip in a stock? Kaiju ETF Advisors CEO Ryan Pannell tells Yahoo Finance Live how his new ETF is using AI to detect dips in stocks.

Video transcript

[AUDIO LOGO]

- Well, we've been talking a lot about the rapid advancement of artificial intelligence lately. And our next guest says he's using the technology, and has recently launched ETF Dip, focusing on buying the dip investment strategy. Joining us now is Ryan Pannell, Kaiju ETF Advisors CEO for this week's ETF report sponsored by Invesco QQQ.

Ryan, it's good to talk to you today. Talk me through how this works. Because, I mean, machine learning has kind of been integrated in ETFs for several years now, right?

RYAN PANNELL: Well, in terms of as a filtering mechanism, if you're looking at the AI's ability to crunch large amounts of data, yes. But in terms of making the endpoint investment decision, no. We haven't really seen that so far. I think there's a gulf of trust that still needs to be crossed when you talk about AI making decisions of consequence on behalf of humans.

We've all played around with ChatGPT, and maybe it's written a funny dinner invite for you or done some light lifting in your business, but would you trust it to write the closing statement in a trial in which you were defending yourself? Would you trust it to write your character reference letter for a job that you desperately wanted? Probably not.

And so when you add an element of jeopardy into the mix, then people's faith in a new technology tends to sort of fade away. And it's going to take a little while to cross that bridge, I think.

- Yeah, and exactly on the legal point, I believe there was just a situation where someone used ChatGPT to write a legal brief, and there were fake cases that were cited in that brief. So just an example of that. So how does this then work in your case, Ryan?

RYAN PANNELL: Well, in our case, we're really leveraging the full capability of AI to do a very specific thing. So areas in which AI is not well-suited, at this point in time, would be, say, managing a global macro fund. It's very difficult for a system that doesn't understand context to estimate what the fallout of a geopolitical decision would be on the regional or global economy.

But when you're talking about dip identification, you're talking about something that's highly technical in nature. And when we talk about dips, we're not just talking about any stock that's sold off, which is sort of the general retail take on buying the dip. We're talking about artificially oversold stocks.

So that is something that has experienced, perhaps, liquidity avoid, it's been predated upon by HFTs, it's been pushed below its near-term fair market value. And there's generally a mean reversion that occurs. And prior to that mean reversion, these stocks exhibit specific behaviors that the AI is able to identify.

It's very short. There's a lot of data that needs to be analyzed in terms of making that determination in a short period of time. It's something that AI is really well-suited to. So that's how this is working. It's predating on those low to high mean reversions, largely in S&P components.

- I mean, what does that behavior look like, and why do you need AI to do it?

RYAN PANNELL: Well, it's going to have a specific volume compression signature, is largely what we're looking for. We look for a number of different criteria, obviously, we're probably not going to share the secret sauce behind it all, but it's specifically quantitative in nature. It's not something that's a result of a bad piece of PR, it's not a result of a miss on an earnings report or something like that. It's literally a systemic mechanical outcome that occurs.

Just very short term. And again, we're not trying to time the market. We're not predicting that this mean reversion will continue in any meaningful way. It's just literally the dip that we're catching. And then we're there for about one to seven days, the average is three.

- Ryan, thank you so much for breaking it down for us. So one of the many interesting ways that people are playing with AI right now. Appreciate it.