I still get a lot of questions about why I teach.
It's always a variation of the question, like, "If you have such a successful formula for investing, why do you want to share it with others?"
For quite a while, I argued that by teaching the subject deeply, I am forced to progressively simplify the concept and understand it based on first principles; this is further honed by serious students with genuine questions who expect answers because they pay top dollar to attend my Masterclass.
Now, with AI and teaching an actual Data Analytics class, things have gotten crazier.
It's no longer about deepening a person's understanding when a lecturer teaches a subject. The lecturer can build something in real time to test the theories they learn and see whether it holds up in reality.
So here's a concrete example.
One of the things my students have to do is linear regression; they might need to figure out which attributes result in better fitness, like lower blood pressure. For every topic I teach, I ask myself how to make it relevant to me and incorporate it into my Portfolio Manager Windows program, which I vibecoded using Claude Code.
So I tried to create a simple regression that looked at just stocks in my portfolio and determined whether a PE ratio and dividend would affect investment gains after a year. And I can see that a low PE ratio resulted in superior performance, but the high dividend stocks in my own portfolio would actually drag down returns after a year.
Upon closer examination, the reason is less dramatic; the outperformers in my portfolio are the AI Tech stocks on SGX like UMS, AEM and Frencken. With the exception of Frencken, I owned these counters when they were still dividend stocks a number of years ago.
Another thing I make my students do is to conduct K-means clustering. It employs an algorithm to divide a data sample into K parts. It's like dividing the class into 5 groups using AI based on the data fed to it, so that insights can be gleaned from this categorisation.
So when I teach this topic, I try to get my students to use their intuition to make a case for how the categorisation is done ( groups like "Fit Daddies" or "Bored Tai Tais" using gym data). It is not an easy task because it's highly subjective.
So after teaching, I began to eat my cooking and tried to cluster the stocks in my portfolio using stock ratios, and it somehow grouped banks with stable REITs, the Tech-based growth stocks, and high-yielding REITs with less prestigious sponsors. More interesting is a cluster of orphan stocks that include First Shipping Trust, Global Investments and, probably unjustified, Capitaland India Trust.
Where analysis can be done, data can be ingested in a built-in AI chatbot.
So, I've gone even further and given my software chatbot capabilities to answer questions about what I should do with my portfolio. In this example, the AI tells me to ditch ESR REIT for Sabana REIT.
My journey continues; this article is to get folks to look out for a future video on whether AI can replace Financial advisors with the good folks of The Financial Coconuts. I suspect the video might not do that well because it is easily one of the most technical talks I've ever done.
