In my first ultralearning project I taught myself the fundamentals of data science and landed my first full-time data science position. The capstone of this project was CupUp, a web app I deployed that used machine learning to calculate the cupping score of a cup of coffee. This was a full-stack data science project that included scraping the web to collect data, cleaning and processing the data, formatting the data to train a machine learning model, tuning the machine learning model, and finally deploying the model in a web app to make its predictions accessible to the web. I was proud of myself for learning the fundamentals of data science and pushing my code to the web, however the outputs of this project left me with more questions than answers. 

The initial intent of building a model that could predict coffee’s desirable flavor notes was to help guide my coffee roasting research process. I wanted to find out which flavors people preferred so that I could look for green beans and roasting techniques that accentuated those flavor notes. This was going to help us create the brand-defining product we could sell for our coffee business. Funnily enough I got to the end of my project before I asked myself whether I could even identify flavor notes in coffee. Needless to say my priorities were in other places. Overall I figured out two things by completing this project:

1)    I have a seriously undeveloped palate and
2)    I strongly preferred coding over roasting coffee.

Our coffee business was struggling to remain profitable and after spending hundreds of hours developing the business and months of trial-and-error coffee research I decided my time would be better invested elsewhere. Specifically I noticed an opportunity that could put my new data science skills to the test in a no-nonsense kind of way that might just end up being lucrative.

Over the next few months I worked with my business partner to transfer the company over to him. Between moving across the country (twice) and switching jobs this took some time. I took a lot of time to reflect on why we struggled so hard make this business successful. We needed to push hard to sell coffee to the point where we went door to door to sell products before they expired. Margins on physical coffee products were razor thin, meaning I could spend hours roasting coffee and barely break even. Don’t even get me started on shipping costs. 
For a long time I wondered what I could do to make our business succeed. Should I have been a better salesperson? Should I have dedicated more time to roasting to make a better product? Should I have reached out to more friends and family to sell additional bags? After turning over these questions for a few weeks I stumbled across a YouTube video by Alex Hormozi that answered this question exactly. Man the YouTube algorithm still knows me a little too well... 

He mentioned a simple yet profound lesson taught by a business professor. The business professor asked their students what they would need to make a successful hotdog stand. The best hotdogs? No, said the professor. The lowest prices? No, the professor said again. The best salesperson behind the stand? Again, the professor said no. After a few more attempts the professor had stumped the students.  Finally the professor gave the answer:

The key to having a successful hotdog stand is to serve a starving crowd.

Bam. That’s the reason that our coffee company was having such a tough time generating revenue. There is no starving crowd when it comes to coffee. People don’t wake up in the morning wondering where in the world they will find coffee when there is a Starbucks on every corner. Even if customers are attracted to artisan coffee (a small subset of coffee drinkers), there is not a strong enough appeal in the United States for our specific niche of artisan Rwandan coffee (an even smaller subset of coffee drinkers). Finding customers who are interested in trying Rwandan coffee is hard enough but when the budget shrinks it’s going to be the first luxury to go. The primary appeal of our product was supporting a local small business, not the coffee we were providing. The next idea I worked on needed to alleviate a pain point that people couldn’t afford to ignore.

I spent the next few weeks looking for strong unaddressed pain points I heard about in daily life. One excellent recurring pain point was difficulty dealing with internet service providers (ISPs). Everyone needs internet and everyone I know hates working with cable companies to get it. This can be for any number of reasons: the service can be slower than advertised, internet plans can be bundled with unwanted products in overpriced packages, and interacting with cable companies is a nightmare. While there is a starving crowd here (disgruntled ISP customers) its solution requires a lot of domain knowledge / startup capital and isn’t something I can help with in the near (~1-2 years) future. I would want to solve a pain point using the skills I have been developing over the past few years.

What would be a good unaddressed pain point that I would have the ability to solve? Well I taught myself data science a couple years ago so maybe it’s worth applying that to a pain point. The goal of data science is to coax valuable insights out of datasets and use those insights to help solve future problems. Knowing this the question becomes: what are insights so valuable that people cannot afford to ignore them?
This is where a lot of my past projects fall by the wayside. Yes they can be a lot of fun to put together, help build foundational skills, and can be personally insightful but I shouldn’t be surprised when I don’t get much more than an “Oh that’s neat!” from others. I need to find insights that can directly impact others in a way they cannot afford to ignore.

Finally I stumbled across the perfect opportunity. I had coworkers who occasionally mentioned sports bets they would place on games they watched. I began learning more about how sports betting worked: what the odds meant, how the bookies make their money, why the odds are the numbers they are, etc. and I found a golden opportunity on two fronts. 

The first opportunity is that sports betting is an information-based pain point, exactly the type of problem that data science was invented to solve. Sports bettors want to place winning bets and to do that they need to know the future outcome of an event. In this case that event is a feature in a game such as how many points one team scored (totals), how many more points that team scored than their opponent (spread), who will win the game (moneyline), or a smorgasbord of other types of bets (props). The amount of data collected on professional sports matches is vast, the outcome that you want to predict is clearly defined by the bet you place, and there is binary feedback about whether or not your prediction was correct. Sports bettors looking for smart bets are a starving crowd.

The second opportunity is that while there is a lot of interest in sports betting, modeling sports bets with excel, and a few sports betting open source projects, I couldn’t find any books on sports betting using python (or really anything much stronger than Excel) on Amazon. I didn’t say I couldn’t find many, I said there weren’t any. In fact while there are plenty of research papers on the subject, I can’t find many python-based sports betting books at all. This may indicate that the niche is too small or it could just be the case that nobody has done it yet. After all, while sports betting with python is a very niche subject it could have an immediate appeal to people who knows how to code but doesn’t necessarily know how to turn that code into cash. A book on sports betting with python could help them bridge that gap.

Here is the punchline: there is an opportunity to use data science to learn how to place smart sports bets using python and publish those results in a book. This will be the subject of my second ultralearning project.

I worked on my own for my first ultralearning project but for this next project I’d like to start building more of a community. This is why I’m streaming myself building this project every weekday from 6am – 8am PST on Twitch: 

twitch.tv/kidbillyy

So please stop by, say hi, and join me on this journey!