I'm a lucky entrepreneur.  The technology company I co-founded in 1991 went public in 2000,  just a few months before the tech bubble burst.  A few years after my retirement the financial crisis hit and the stock market collapsed.  I'm not a lucky investor.  So I've built a learning machine to help guide me.

 

Call me skeptical, even cynical.  Experience has taught me a lot can go wrong in the world.  When the tech bubble burst, our company's stock price collapsed along with the entire tech sector.  Although profitable and growing we became one of the "babies" thrown out with the bath water.  It was apparent then that macro-trends can overwhelm individual valuations.  The company survived, kept growing and 4 years later was acquired.

 

Not long after I retired, the financial crisis occurred.  The stock market lost almost half its value.  The experts that professionally managed my investments had few satisfying answers.  They had provided me with their suggested portfolio of funds and it suffered a significant loss.

 

Given those sad realities, I began a deep dive to better understand broad market behavior.  Being retired and a "techie" I've had the time, the data and the analytical tools to explore my curiosity.  That lead to a study of behavioral finance, machine learning, and artificial intelligence.  From there I built a primitive learning machine to help me with my own investments.  A patent on some of my original methods yielded a web site, InvestEngines, others could use. Soon, the mountain of data the machines collected was overwhelming.  It became too complicated and not easily exploited as a result. I had beaten a fly to death with a sledge hammer.  And what started out as a fun hobbie, was beginning to morph into a potential business, which was more work than fun. So I shut it down and went back to the drawing board.

 

Most people don't care for analytics and arcane explanations.  Friends and family showed an interest in InvestEngines until I started explaining the details and their eyes glazed over.  The financial professionals that used InvestEngines were uncertain about what its indicators were indicating.  Everyone just wanted simple and actionable insights that made investing choices clearer.  The good news was that the learning machines had uncovered some important relationships in the market data.  Since then InvestEngines has evolved significantly.  This blog is a simple view into its conclusions and how they invite a better investing strategy.  It is a distillation of what the learning machines have taught me along the way and what they are teaching me now.   

 

InvestEngines has become my GPS for selecting the best route to a financial destination.  The articles that follow offer simple suggestions and important insights.  I will post actionable signals and quarterly updates on the "latest" tab.  There is a lot of detailed data and analysis behind the conclusions.  You don’t have to read it all to get the gist, but it's all there for context and clarity.  To make it as simple and accessible as possible, I start with the basic concepts in Article 1, progress to more details and don't get deeper into the weeds and models, until Article 7. Beyond that the Articles become more current as the models evolve.

 

The world is changing and this is just a start, just a prototype of what may come.  It is clear to me that artificial intelligence and machine learning have a growing role to play for all investors.  The learning machines keep learning and so do I.  InvestEngines is free for anyone to use, because this is one of my retirement hobbies and I use it myself.  But a word of caution; it may only be worth what you pay for it. 

​

Neal

2019

​

You can contact me at investengines@gmail.com

​

August 3, 2024 thoughts

lnvestEngines has been running live for almost 3 years now and you can see all the updates in the"latest" tab. The models are still using only the training data that ended in 2021. The recent outcomes still fall within the range of the prior 20 year training window.  The 30 month performance of the machines was a good test as it included the negative impact of the pandemic. The value of the 60/40 benchmark is pretty much where it was back then. During that same period the learning machines have managed to keep average annual returns to almost 8%; all done with only a handful of re-allocations. The August 2024 update shows the previous 12 month’s InvestEngines’ return was almost twice that of the benchmark 60/40 allocation.

But there is plenty to be skeptical about.  InvestEngines' is still primitive relative to the rapid evolution of AI, especially in light of the recent emergence of Large Language Models (LLMs). Things change rapidly. The cyclicality of the market is definitely accelerating, its noisy data, and the training period did not include circumstances that precisely mirror today’s. Tipping points can happen suddenly, especially in periods of uncertainty following periods of stability. So no promises, but my hobby is to try to improve the models. Stay tuned.

Neal