March Madness Modeling?

Greetings Programs!

I’m generally a beginner when it comes to R(Studio) and brand new when it comes to python, but I am a jedi master(and I sit on the council) when it comes to sports fandom, in particular march madness when my team is good(they are this year). Additionally, I actually like my coworkers and enjoy finding ways to engage with them other than “it’s not ready yet” and “can you help me with this thing” or even “Yeah, I’ve got a few minutes, maybe I can help with your problem”

Are there any websites out that there than can help guide you through using either Python or R(Studio) for running models to fill out your march madness bracket?

Best case scenario there is a resource out there which can show you where to pull historical data, either form existing csvs or tab separated files, or newb friendly web scraping(and be accessible at work so I can then share with my coworkers)

After the data is downloaded to our machine, there’s some guide work for how to build the model, including example or actual syntax to use in either R or Python to begin to manipulate the model. I’m not sure something exists that can predict each round, but that would be fantastic. Then, and here’s where it’s fun, guidance on how to tweak the model in your software, so you can test different assumptions and try different approaches.

Ideally, this teaches how to actually execute some of the code, gives a little familiarity with the tool, a little familiarity with “data science” and can support friendly competition for the best bracket.

I tried following along here, An R Shiny App to Streamline March Madness Predictions | by Sabi Horvat | Towards Data Science , but work unsurprisingly won’t let me access sports betting websites

https://www.sportsbookreviewsonline.com/scoresoddsarchives/ncaabasketball/ncaabasketballoddsarchives.htm

Has anyone followed through with the 538 approach themselves?

Till all are one,

Epistemus

Kenpom is great. Kenpom is the best. There’s all sorts of data on teams, which can give a probabalistic model for any matchup. From that you can expand with monte carlo simulation to your little heart’s desire.

The subscription is $25 per year.

Never took Deep Purple for a bot.

2 Likes

That’s cute.

Just to be clear, Kenpom is chock full of basketball statistics and data from games. There’s no programming help or such offered. Just summary stats. Im just an actuary that admires it because it’s basically an avalanche full of data on sports. And I don’t even gamble. Seriously, never.