March Madness is almost here!!! Let's walk through building a predictive model for the tourney using the CBBD Python package and XGBoost. - ThreadSky

About ThreadSky

collegefootballdata.com • 10 days ago

March Madness is almost here!!!

Let's walk through building a predictive model for the tourney using the CBBD Python package and XGBoost.

Comments

joeprogrammer.bsky.social•5 days ago

Thanks for publishing this post. Very interesting! Straightforward and easy to understand. I even got the code to work the first time! I'll use this to make a bracket this year. It should beat my usual "Who would win a fight between a bulldog and a wildcat method". 😆

joeprogrammer.bsky.social•5 days ago

Do you think the model is baking in a home court advantage? Running your example I get Michigan over Dayton by 9.00, but flipping the teams in the function call (Dayton home, Michigan away) gives me Michigan by 2.37. For neutral court tournament games would we split the difference (Mich by 5.6)? 🤔

collegefootballdata.com•5 days ago

Good question. I believe the training data generally had the higher seed listed first so it's probably baking that in and I would say to ensure best results to also follow that convention with predictions.

collegefootballdata.com•5 days ago

An interesting experiment would be to shuffle up the training data and sometimes list the lower seed first. I bet that would even out a bit.

joeprogrammer.bsky.social•5 days ago

Ah. Gotcha. Ok, thanks for the response! GL with your bracket!

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply