Making Sports Predictions with Data Science

🏆 WIN a NVIDIA GeForce RTX 4080 Super GPU! Register now: forms.gle/47vUHzzz2aqJoP1S9
Dive into the fascinating world of machine learning and AI as we guide you through developing a model designed to predict NCAA tournament outcomes. From initial setup to final predictions, we’ll cover everything you need to create your own powerhouse model.
Embark on this machine learning adventure to not only enhance your NCAA bracket but also to deepen your understanding of predictive modeling. Let’s transform your bracket into a data-driven masterpiece together!
📘 Access the Notebook: www.kaggle.com/robikscube/mac...
Timeline:
00:00 Intro
01:06 Setup and Import
02:39 Data Understanding
05:23 Data Pipeline Goal
06:04 Step 1- Season Data
11:07 Step 2- Tourney Data
15:30 Chalk Baseline
17:51 XGBoost ML Model
21:30 Predicting Matchups
23:50 Creating the Bracket
Links to my stuff:
* KZbin: youtube.com/@robmulla?sub_con...
* Discord: / discord
* Twitch: / robcodeslive
* Twitter: / rob_mulla
* Kaggle: www.kaggle.com/robikscube

[@TTOmiTTom]

Rob please don't stop making videos. I just discovered your channel and my universe just expanded!!! Awesome stuff!

[@elu1]

The pre-processing was mind boggling to me. will have to go through it with the notebook. Thanks Rob.

[@robmulla]

Yea. The data pipeline is definitely the most complicated part of the entire process. But if you can learn a lot by trying to understand it!

[@buccaneerbulldog192]

Such a fun watch. Really enjoy the sports models.

[@chillphil967]

that was sweet 🤙 granted, i’ll have to come back and hit pause at every step of the way again lol. but, good content and great level of depth.

[@ThePaintingpeter]

I'm not into the tourney itself, but your code presentation is always insightful. Very much appreciated. 😂I'm sending you a fiver to improve my winning odds for the GPU :)

[@chacehawkins4708]

Hi Rob, I love your videos. You have saved my butt more than I can count, and I am using your videos to help me move up in my career and stay interested in learning more complicated tools. Would you ever consider hosting a discord channel for data analysts, or do you know of any good ones. (or does anyone reading comments know of one). A place to ask questions would be AWSOME!

[@ed-salinas-97]

Really nice work. Thanks for sharing. One issue that might come up, though, I don't think 538 does sports rankings/ratings/predictions anymore. Last year was the final year of that, so there may not be any new data for 2024. I could be wrong, but since I typically use their data for baseball and NFL games, it was a huge hit to me when they decided not to do it anymore last summer.

[@robmulla]

Thanks for mentioning this! I did not realize they were no longer creating sports forecasts. In that case the 538 features can’t we used in this years model- but it may cause people to create more creative features on their own!

[@Break_down1]

Just found your channel. This is great content !

[@robmulla]

Welcome aboard!

[@dimitarbogoev4539]

hes back!

[@antonio613]

I love your Videos! Amazing python learning.

[@robmulla]

Happy to hear that!

[@user-oj1py3dn1b]

Your videos are so amazing, can you make some videos about deep learning model

[@robmulla]

I appreciate the comment. Have you watched my videos on sentiment analysis and intro to pytorch? Those involve deep learning.

[@XEQUTE]

Hey can you do some finance kaggle competitions, like the home credit risk @@robmulla , I hear quite some drama when the competition got metric hacked and they started it again based on a new stability metric

[@gyanajyotipradhan4519]

I have one question reg large dataframe(3gb) created after melting two dataframes in pandas. How can we save the df, it's showing memory error

[@twoodcc]

i just found your channel. i tried to run the model, but i get several errors, like this below. i'm new to kaggle, so maybe i'm missing something: ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. tpot 0.12.1 requires xgboost>=1.1.0, which is not installed. keras-nlp 0.8.1 requires keras-core, which is not installed.

[@robmulla]

Try commenting out or deleting all the cells up until the main imports. There seems to be an issue with the kaggle notebooks.

[@SearchingforScraps]

would i be able to do this on a jupyter notebook?

[@robmulla]

Yes! Kaggle notebook is essentially a Jupyter notebook.

[@quintonrichards4805]

So how would I update this (as someone totally new to ML) to reflect the current brackets that were released?

[@robmulla]

The Kaggle folks still need to update the dataset with the picks. Also 538 is no longer releasing predictions unfortunately. So you would need to remove that feature from the model.

[@quintonrichards4805]

Thank you so much! I’ll go in and see if I can do that!

[@rathgrams]

I like learning about data science from you, but also how do you get your hair like that please?

[@robmulla]

Haha. Thanks! I just use a comb and fight with it.

[@rathgrams]

Hurray I got a reply. Dam that’s it. I’ll try that thanks

[@swiatlowiekuiste]

Of course I'm learning ML for the sole purpose of making people jelaous 🤭

[@XEQUTE]

nvidia GTC , rob mulla , LFG!!

[@ericpayne2846]

Where have you been for the last 5 months ??? !!! 🏓

[@robmulla]

I’ve been hereB

[@Superteastain]

Anyone tried this on a Jetson Nano? I'm going to kick the tires and see if it will run

[@thomaspelling1399]

There's no way Rob Mulla will ever be ballin...

[@Superteastain]

I think its totally a generational thing but man I can read black on white easier than white on black.

[@robmulla]

Thanks for the feedback. Be curious to hear if others agree.

[@Superteastain]

Your follow alongs are really good. @@robmulla

[@henk_iii]

Drop down and give me 10

[@fdepierre]

Merci !

[@TheEditingAddict]

Cool tutorial, but the model just predicts the lower seed to win every game lol

[@robmulla]

Did you watch it all? I compare the model performance against the lower seed only predictions.

[@brianwhite3428]

Here's some advice Why not just show? Who AI thinks will win the games Like a AI can do any better? LMAO!!!!!!!!!!!!

[@robmulla]

Thanks for the advice. I don’t understand what you are trying to say though.

[@brianwhite3428]

@@robmulla Who did AI pick in all 63 games to win? It's that simple Did AI predict any upsets?

[@robmulla]

@@brianwhite3428 oh. When I made the video the field wasn’t set yet. This video shows you how to train your own model with your own features and ideas- not to just give you the picks. Sorry!