I've had a good run over the past couple years from simply picking off bad lines. A good enough run that I've been able to live comfortably without any other sources of income. While a profitable strategy, I sense that the window of opportunity may be slimming down or even coming to an end. An increasing number of books are able to mirror Pinny or BM in real time, and so I feel this may be an unsustainable way to generate positive yields going forward.

With that in mind, I think it will be useful to learn some new skills and techniques to try and arrive at the correct lines independently. While I understand many of these markets may be already solved by people and groups with years or decades of experience, I still think it will be a fun journey, and who knows? Maybe I'll be able to find a niche prop or two that have some opportunities.

What I'm looking for is guidance from people who originate their own lines. What skills have you found to be most important?

I don't have much experience here, I'm an above novice excel user, familiar with things like VLOOKUP, pivot tables, and arrays. But I want to learn to scrape and use play-by-play data, and I hear excel isn't very useful for large data sets. I'm looking at codecademy and considering starting with this one, related to python and data science:



Is this a reasonable starting point? What other skills and applications might have value when it comes to creating one's own lines? Thank you to anyone who had the time to read this and can help point me in the right direction.

Math/tech knowledge necessary to build a worthwhile predictive model

Ignoring the Excel question, the real question is what are you going to do with the data once you've scraped it? It's all well and good saying there's a 100% correlation between the number of quarterback kneels taken and the number of games won, but good luck with that if an offense ran that every play

Problem with Excel and big data is the limitation on sheet size. If you dump things into an SQL table you don't have that issue. But that's secondary - my "data model" is just one CSV linked into three measures in Power BI, then again the sport I bet is incredibly numerical and easy to model as a result. It's what you model that's important.

I'd definitely start by reading TomG's link.

As someone who followed the advice in that thread and is now an analyst at a sportsbook, I'd suggest one change: start with Introduction to Statistical Learning. It was written after Elements of Statistical Learning and is easier to understand and apply the techniques described. It will also introduce you to the statistical programming language R to see if you like it and I think you can still find it for free. An even easier/faster read is Statistical Sports Models in Excel vols 1 & 2. That guy wrote it while he was a data science student and you probably won't learn enough to build winning models, but I think those excel books cover the absolute minimum you should know to at least start originating.

I get tasked with a wide variety of predictive analytics and modeling problems, and mostly use Python and SQL on the job. I'm familiar with the main classification, regression, and visualization libraries in R, but overall I prefer Python because I find it easier to do non-analytical tasks such as web scraping with. Pinnacle's team uses mostly R I've heard though.

Still, I get the impression low-code/no-code analysis apps are becoming more popular, so taking a Python or R data analytics course like the one you screenshotted is recommended but not the only option. Excel is still a great tool as long as the size of your dataset fits into the row/column limits. Just over 1m rows which should be plenty if you are just starting to scrape and build a database anyway.

Onshore or offshore?

I've only worked onshore regulated books so far. It's been the same situation for the few industry people I've bumped into on this forum. If anyone with offshore experience reads this: DM me, always down for some networking.