The Journey

Why build stockrobot.ai?

It was August of 2020 - in the heart of the COVID-19 epidemic. Many recent market dislocations had creating some serious trading opportunities (I was detailing my trades and thinking on my old blog at the time). Most of the time I'm actually a lazy, pseudo-efficient-market hypothesis guy. Most investors should not try to pick stocks and actively manage their portfolios. Active investing is hard, risky work. My typical advice for friends and family where I'm not actively managing their investments is to stay in some mix of SPY (the SP500 index) and bonds. Standard finance 101 efficient frontier portfolio stuff.

But that advice leaves a lot to be desired. Even if someone has resigned themselves to passive investing via an index fund like SPY - there are still questions about when to invest. The market can be volatile - should I wait until a sell-off to add to my SPY position or just blindly dollar cost average as cash becomes available? Should I ever consider selling the SPY to limit my exposure?

I honestly never knew how to handle these topics. After years of investing I've developed a loose sense when things are stretched to the up or downside, which is how I (used to) make my entry / exit decisions. Being a data scientist I found this intuition-driven approach totally unsatisfactory - even if historically effective. And worse, it's impossible to communicate to someone else - a totally opaque method for making decisions. I may as well be reading tea-leaves.

So I started out thinking about how I could solve this problem. Technical analysis has never been my cup-of-tea. Most traders doing technical analysis are using the same indicators, doing the same backtests, and arriving at more or less the same strategies. It all felt like pure survivorship bias and hype. Importantly, there wasn't any theoretical framework in which to root my thinking on technical analysis - the same indicators could imply "momentum" and indicate a buy signal, or "overbought conditions" indicating it's time to sell. It all felt... arbitrary. And indeed I think most of technical analysis is exactly that.

What I wanted was a way to explore all technical indicators that would evolve over time, and be constantly tested against the markets. I wanted to have an engine for solving the tactical question of "When should I buy / sell?" to support my fundamental analysis.

And thus stockrobot.ai was born.

The conceptual framework

The concepts behind stockrobot.ai have been baking in my mind for 20 years.

In 2002 Dr. Stephan Wolfram published 'A New Kind of Science' (NKS) in which he documents the key principles and concepts he's discovered while exploring the computational universe. As a young physics / philosophy undergrad I was disenchanted with the lack of exciting, paradigm shifting ideas in physics. I didn't want to make predictions, I wanted to understand physics at its core: and in Dr. Wolfram's I sensed that potential. I ripped through all 1200 pages in a few weeks and became enchanted with the idea that computation was at the heart of everything.

After graduating from my MBA in finance 7 years later I was at impasse. I'd been optimizing logistics systems at UPS after my undergrad and was managing non-trivial portfolios for friends and family. UPS had run its course, it was a great job but not the career I imagined for myself.

At the time I thought I wanted to run a hedge fund, and getting my MBA would offer a gateway into that world. The financial crisis of 2007-2008 had other ideas. I found myself applying to positions at investing funds and my CV going head-to-head against recently unemployed Bear Sterns fund managers with 30 years of experience, willing to work for a pittance. I needed to find another path forward.

In my free time I'd be learning how to program. I was attempting to build computational models of economics to explain how culture and social trust interplayed with legally enforced rights to produce conditions of economic growth. The goal was to model exchange at the lowest level of many interacting economic agents to see how their emergent behaviors compared to the prevailing economic framework expectations. It was basically stockrobot.ai for economics.

I was leaning heavily on the ideas in NKS for the work. One night while researching I happened upon the NKS Summer School - a school put on by Dr. Wolfram meant to help explore NKS based projects.

On a whim I sent through my manuscript. To my amazement I was accepted. A few months later I was walking around the leaning tower of Pisa, talking to Stephan Wolfram about his latest project: Wolfram Alpha.

The Plan

I would go on to leave my job at UPS and join Wolfram Research as a Principal Consultant, and later the Director of Technology for the consulting group Wolfram Solutions. In this formative period I was getting applied experience leveraging machine learning and artificial intelligence techniques to solve real-world problems. It was an amazing experience.

After designing my own version of the then-popular random forrest ML method (libraries for all of these things were fairly nascent, Mathematica's powerful Predict and Classify functions had yet to be created), I started to using it make predictions in sports. Daily Fantasy Sports (DFS) was just coming into its own and having access to sophisticated models provided an edge, allowing me to make a nice side income while exploring ML applications on projects of personal interest. I wrote yet another old blog to document my process and picks. This blog was discovered and launched an 8 year adventure in the world of sports betting, and several successful company / product launches.

Which brings us to today. I love my work at Hard Rock Digital and do not intend on leaving any time soon. For now - stockrobot.ai is a side project / running experiment. On nights and weekends my founders and I use stockrobot.ai as a project to explore new technologies, interact with and build novel communities, and follow the model of the Minimalist Entrepreneur.

Over the longer term I aspire to make stockrobot.ai into a game, complete with user-owned and managed pools of bots, user-driven evolution decisions, and user-managed subscriptions to bot alerts (with the users being able to market and profit off their most profitable bots). For now - we're opening up just the basics to our early adopters. Looking for those who believe in the potential of this approach and are interested in gaining access to one-of-a-kind evolutionary trading signal platform.

So join us. Sign up for a membership through our Discord server - for just $5 a month you'll get access to all bot signals, and a monthly newsletter outlining the development of the platform and deriving insights from the bot ecosystem.

The Journey

Why build stockrobot.ai?

The conceptual framework

The Plan

This website uses cookies.