MARCH MADNESS - Will My Machine Learning Model Beat Your Bracket?

Подписаться 260 тыс.

Просмотров 19 тыс.

50% 1

Will my machine learning model dominate all other March Madness Brackets, or will my friend Bobby win all of my most important belongings? Find out in this new video!
Model comparison tool: share.streamlit.io/playingnum...
Simulation Tool: share.streamlit.io/playingnum...
Kaggle Notebook: www.kaggle.com/kenjee/simple-...
Elo Rating Website: warrennolan.com/basketball/202...
Follow my Bracket: picks.cbssports.com/college-b...
I haven't done any sports analytics content in a while, so I thought it would be fun to make this video about how I used machine learning to fill out my NCAA Basketball tournament bracket.
I walk you through the 5 different models that I use. 2 machine learning based, 1 simulation based, 1 Elo rating based, and 1 based purely off of seed. I also made some pretty cool tools that can help you see how everything works under the hood.
Hopefully next year I will finish my analysis before the tournament actually starts so you can use the tools that I build.
#DataScience #KenJee #SportsAnalytics #MachineLearning #MarchMadness
0:00 The Challenge
0:38 Intro
1:03 What is March Madness?
2:00 The First Day
2:36 Monte Carlo Sim
3:34 Model 1 Results
4:22 Reset
4:55 Hack Day 2
5:20 The 5 Models
6:00 Random forest and Neural Net
6:45 ELO Rating
7:35 Seed Differential
7:45 Aggregate model
8:15 My Bracket
⭕ Subscribe: ru-vid.com?sub...
🎙 Listen to My Podcast: / kensnearestneighborspo...
🕸 Check out My Website - kennethjee.com/
✍️Sign up for My Newsletter - www.kennethjee.com/newsletter
📚 Books and Products I use - www.amazon.com/shop/kenjee (affiliate link)
Partners & Affiliates
🌟 365 Data Science - Courses ( 47% Annual Discount): 365datascience.pxf.io/P0jbBY
🌟 Interview Query - www.interviewquery.com/?ref=k...
MORE DATA SCIENCE CONTENT HERE:
🐤My Twitter - / kenjee_ds
👔 LinkedIn - / kenjee
📈 Kaggle - www.kaggle.com/kenjee
📑 Medium Articles - / kenneth.b.jee
💻 Github - github.com/PlayingNumbers
🏀 My Sports Blog -www.playingnumbers.com
Check These Videos Out Next!
My Leaderboard Project: • I Built the FIRST EVER...
66 Days of Data: • What is the #66DaysOfD...
How I Would Learn Data Science in 2021: • How I Would Learn Data...
My Playlists
Data Science Beginners: • Data Science Beginners
Project From Scratch: • Data Science Project f...
Kaggle Projects: • Kaggle Projects

Спорт

Опубликовано:

5 июн 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 108

@KenJee_ds 3 года назад

Thanks for watching everyone. All links I mentioned are below 👀 Kaggle notebook for sim & ELO: www.kaggle.com/kenjee/simple-simulation-elo-rating-approach See if I win my 50 Papayas here: picks.cbssports.com/college-basketball/ncaa-tournament/bracket/pools/kbxw63b2gu4dombrhezq====?entryId=ivxhi4tzhi4dqnjsgm3dkna= Model comparison tool: share.streamlit.io/playingnumbers/march_madness_predictions Simulation Tool: share.streamlit.io/playingnumbers/basketball_sim_dash/main Elo Rating Website: warrennolan.com/basketball/2021/elochess

@timothykenneth4925 3 года назад

Dont know if you guys cares but if you are stoned like me during the covid times then you can watch pretty much all of the latest movies and series on instaflixxer. Been binge watching with my gf lately xD

@mauricecasen7302 3 года назад

@Timothy Kenneth Yea, I've been watching on instaflixxer for months myself :)

@LukeBarousse 3 года назад

Heck yeah!! Back to the sports content, loved that you took us through the journey of building this model, awesome to see this perspective!

@KenJee_ds 3 года назад

Thanks for watching Luke! It was far past due!

@DannyMa 3 года назад

Good to see you back in the sports analytics game! 🔥

@KenJee_ds 3 года назад

More to come!

@johnginos6520 3 года назад

This is great content if you wouldn’t mind. Sharing more about the model averaging and stacking in sports analytics? Is this normal practice what are the most common models for sports analytics... say-basketball in particular ?

@KenJee_ds 3 года назад

Thanks for watching John! This is a very common method for basketball. Usually people will base something like this off of the "4 factors model" and a combination of the other rankings systems (sangrin, kenpom, etc.).

@danielsilva3383 3 года назад

Hi great video! Sports analytics is definitely cool.

@KenJee_ds 3 года назад

Glad you think so! Thanks for watching!

@Ibraheem_ElAnsari 3 года назад

I know very little about Sports in the US but this seems to be very interesting, as were the DS concepts you shared .. I really hope U win 🤜

@KenJee_ds 3 года назад

Glad you found it interesting Ibraheem!! Thanks for watching!

@thinamG 3 года назад

I love sports. This is amazing, Ken.

@KenJee_ds 3 года назад

Glad you enjoyed it Thinam!

@DannyMa 3 года назад

That live-edge dining table at 5:10 is also LIT as 🔥

@KenJee_ds 3 года назад

Yeah it is sick, may try to steal it from my friend 😂

@isaactrussell5121 3 года назад

Ken! I made a dataset on Kaggle with all the NBA top shot nfts. I would be curious to see you work your sports analysis magic on it and see if you could find some correlations between player performance in the season and the value of the players moments on the marketplace!

@KenJee_ds 3 года назад

That is awesome! I'm very interested in NFT's! I actually have a podcast guest to come in and talk on them in 2 weeks. Will check it out if I have some extra time this week!

@jamesscouten2966 3 года назад

Quick story......I live in Vegas. I love sports. Obviously, it was only a matter of time before I started thinking I could "beat the book." Using my brain and spreadsheets wasn't cutting the mustard, so I started researching how people used computers to figure out more efficient ways of being a winning player. That's when I discovered data science. I then saw how much more to the field there was than just figuring out how to beat Vegas. I was hooked. And that is how my data science journey began.

@KenJee_ds 3 года назад

That is awesome! Actually not that far off from my origin story! I wanted to figure out how to predict golf outcomes (for draftkings)... and look at me now I guess

@wilsonman8661 3 года назад

Oh man, not fifty papayas. That's a lot on the line haha Good reminder to have a simple baseline model to compare to; simplicity is important! Also recognize the Monte Carlo Simulation from the GS/Cleveland video from way back. :D

@KenJee_ds 3 года назад

This is Very serious Wilson! And yes, major throwback alert!

@fknight 3 года назад

Man is that your place at 5:05?? Got the vibes!

@KenJee_ds 3 года назад

Unfortunately not, it is my friends haha

@fknight 3 года назад

@@KenJee_ds Well his place has got the vibes! haha

@KenJee_ds 3 года назад

@@fknight I know, it is so sick!

@TheOriginalTakarin 3 года назад

Ok, this is the 3rd time in 2 days I've seen / heard Rothman's book be referenced. Time to buy it.

@KenJee_ds 3 года назад

💪

@cameronyoung3067 3 года назад

Do you plan on doing more NBA game simulation videos? Like incorporating not only previous scores from games, but player stats and trends as well? Found it super interesting and made some additions myself

@KenJee_ds 3 года назад

I would like to! Hoping to do more sports content this year. I might do a retro for how this project went as well!

@ed-salinas-97 3 года назад

I'm curious about the Stacking you mentioned...to find out what to weight a number of models. Is that using something like StackingRegressor in sci-kit learn for example?

@KenJee_ds 3 года назад

It is exactly that!

@piyushzope10 3 года назад

Thanks for the new sports analytics video. I came across many new terms and concepts through this video. Wish you very good luck with the march madness... #PapayaKing

@KenJee_ds 3 года назад

Glad you liked it! hopefully I will become the king!

@piyushzope10 3 года назад

@@KenJee_ds Yepppp

@shrutijain1628 3 года назад

I am not that into sports .. but trying understand everything you said in this video haha🤣💯🙌

@KenJee_ds 3 года назад

Thanks for watching regardless Shruti!

@leavonfletcher4197 3 года назад

I....didn't even know this was possible! And a Kaggle competition too? Next year I am doing this. And I need to do some sort of algorithm for football season too! Awwwww man. I'm hooked.

@KenJee_ds 3 года назад

Do it!!!

@leavonfletcher4197 3 года назад

@@KenJee_ds I am! That sounds amazing!

@asadon1950 3 года назад

This is one of the nerdiest and funnest videos I've ever seen.

@KenJee_ds 3 года назад

Haha thanks!! Really makes me happy to hear this!

@mnm8818 2 года назад

my rant: I think more important ones are matchups - eg one team can win all games all the time vs another team but loose every game to other teams + home court advantage - think betting odds already do this already? and they have a big team of coders/ prob analyst I'm guessing.

@KenJee_ds 2 года назад

They likely are!

@itscrewston6684 Год назад

A data science/analytics channel that works with basketball data?! Hell yeah

@KenJee_ds Год назад

Trying to do more projects later in the year!

@bhuvaneshwariraman1031 3 года назад

All the best!

@KenJee_ds 3 года назад

Thanks!

@JBB685 3 года назад

As an Auburn fan, that Virginia vest gives me flashbacks

@KenJee_ds 3 года назад

Lol. I had to live with them being the first 1 seed to lose in the first round for a full year..

@rithwikraman188 3 года назад

Yoo I see you're wearing a Virginia sweater - are you an alumni? I'll be starting there as a CS major this coming fall :)

@KenJee_ds 3 года назад

That's awesome! Yes I did my first round of grad school there. Awesome place!

@ethanvanlerberghe6425 3 года назад

Great video! Except the link to the bracket isn’t working- I believe you posted the link to join your group instead of a link to a viewable bracket

@KenJee_ds 3 года назад

Thanks for note! Will change it now!

@krishnabarfiwala5766 3 года назад

Hi, I know basics about the models. Now I want to know how and when to use each model like DECISION TREE, CLUSTERING, SVM, K MEANS, K-NN, RANDOM FOREST, NAIVES BAIYES, ETC. I have modeled them all once but I am not sure what kind of interview questions will come with these techniques. PLEASEEEEE PLEASE REPLY TO THIS QUESTION. hey PLEASEEEE RESPOND, LOVE YOUR CHANELLL AND FOLLOWING IT TOOOOOOOOOOOOOOOOOOOO.

@KenJee_ds 3 года назад

Thanks for watching my channel Krishna! Unfortunately, the best way to learn when to use each model is to work on a lot of projects to understand them. I generally recommend doing a few projects where you use all of them and see which performs the best. After you see which performs the best for that scenario you can understand why. I hope this helps!

@krishnabarfiwala5766 3 года назад

@@KenJee_ds okay so I'm every sum I should model each method??

@danielsilva3383 3 года назад

Hi, Great Video! Tell me something: I am building the tennis analyzer system that uses object detection and pose estimation and I would like to know if I should make an web app or a mobile app? What would be more efficient in terms of cost? Thanks

@KenJee_ds 3 года назад

I would try to productionize it with streamlit! You could get a working prototype up for free!

@danielsilva3383 3 года назад

@@KenJee_ds Thanks! Does it accept payments, in case I transform it into a SaaS?

@WingfanSports 3 года назад

Quality video once again. Sign me up for all the sports analytics content.

@KenJee_ds 3 года назад

Haha more to come!

@averydrago 3 года назад

Hi Ken! So I am a Stats undergrad because I want to get into AI. Should I stick with Stats or go to Comp Sci? Thoughts, ideas, opinions? Thanks!

@KenJee_ds 3 года назад

I think it is fine to stick with stats! Maybe pick up a minor in CS if you still have time!

@averydrago 3 года назад

@@KenJee_ds Yea, I still have time. I am a freshman. Thanks for the reply!

@DataProfessor 3 года назад

Great video Ken! 1:04 is that a real papaya?

@KenJee_ds 3 года назад

Haha just art

@DataProfessor 3 года назад

@@KenJee_ds Wow, it looks so real though 😆

@KenJee_ds 3 года назад

@@DataProfessor Haha I wish it was a real one... I'm hungry

@danielsilva3383 3 года назад

Hi what do you think about college? Do you think that project based learning is the best way to learn or is it better to learn concepts? Thanks

@KenJee_ds 3 года назад

I prefer project based learning!

@danielsilva3383 3 года назад

@@KenJee_ds Great! Me too

@AJ-ly8si Год назад

Can you update the data and make one for 2023 once the bracket is released?

@Anthony-Webb 3 года назад

I am not gonna lie, I thought you had said Pot Pie's in the beginning. I was thinking that was a bet I could get behind.

@KenJee_ds 3 года назад

I also love a good pot pie! Maybe I will pay bobby in those if I lose. Although, he did have UNC going to Championship game so I'm not looking too bad

@victorpinasarnault9135 3 года назад

Lke 91º Thank you man! ;)

@KenJee_ds 3 года назад

Thanks for watching!

@pastorsoto1298 3 года назад

Do you think that building an Elo ranking might be appropriate for golf? I am trying to build a golf model but I am somewhat lost.

@KenJee_ds 3 года назад

I'm actually in the process of doing just that! Not sure if I will make a video on it or not though haha

@pastorsoto1298 3 года назад

@@KenJee_ds Would love to see it! Any Idea what to read to build my own Elo ranking for golf? I am not sure if it is my idea but golf seems to be very different than the rest of sports to do data science

@piyushthakur2386 3 года назад

I see Ross in you. Organized , love dinosaurs and tidy hair. No doubt Ross Geller 😎. Jokes apart, I am also brain storming to find my core for Data Science domain selection.

@KenJee_ds 3 года назад

Haha, I never was a big friends guy. The key is to experiment!

@TinaHuang1 3 года назад

The best!

@KenJee_ds 3 года назад

Actually it is in place 204 out of 700.

@aamnasuhail4468 3 года назад

I didn't understand anything. Still watched the whole video. I don't know why 😂😂😂.

@KenJee_ds 3 года назад

Thanks for watching! Hopefully you still enjoyed it :)

@aamnasuhail4468 3 года назад

@@KenJee_ds Yeah, I did. I'll comeback to it when I will learn more things. So, that I can understand it .

@zarkali7164 3 года назад

Yes you can.

@KenJee_ds 3 года назад

💪

@harshitsati 2 года назад

Betting on all of your papayas? Damn

@KenJee_ds 2 года назад

Luckily I won!

@AdriaanVanNiekerk 3 года назад

I miss the T-Rex 🦖

@KenJee_ds 3 года назад

He's still patrolling my other office!

@AdriaanVanNiekerk 3 года назад

@@KenJee_ds ever thought of getting a Pteranodon?

@aaronhess 3 года назад

8:28- lol they lost in the FIRST round

@KenJee_ds 3 года назад

I shouldn't have liked this, but yes lol

@datalyfe5386 3 года назад

I wonder how many people picked Drake to win just because of the name

@KenJee_ds 3 года назад

Probably too many lol

@josephstowell1995 3 года назад

Lmao I didn't sleep for the last 48 hours to build an ML model and it gave me a bracket thats in 4th to last place in my pool smh

@KenJee_ds 3 года назад

Haha mine isn't exactly doing great either. A lot of big time upsets today!

@josephstowell1995 3 года назад

@@KenJee_ds hehe well it got everything right today so thats a plus!

@aamnasuhail4468 3 года назад

And what's the deal with Papayas. Do you love papayas so much ?😅😅

@KenJee_ds 3 года назад

It is one of the worlds greatest mysteries!

@aamnasuhail4468 3 года назад

@@KenJee_ds Hahaha..... okay

@Raiseren 2 года назад

Did you backtest anything? Why did you opt for a simulation model when predicting game winner without having in-play data? It honestly looks very random what you did with little thoughts to your strategy. This is something I would expect from a very junior data scientist. Not someone having a youtube channel intended to teach data science.

@KenJee_ds 2 года назад

I tried to make this more of an intro video and didn't show much of the coding portion. This is a link to the work I did with Hunter this year if you're interested: medium.com/@HunterKempf/predicting-the-2022-ncaa-basketball-tournament-using-data-science-3400a2c84098

@Raiseren Год назад

@@KenJee_ds Thanks. I think I see what might have happened. Generally speaking, Elo rating difference should be the most essential feature. The fact that it is not suggests something fishy is going on (leakage probably). If I had to guess a feature like RPI probably uses future data. You should only look at the data available before the match starts. Whereas I think the RPI is done post-season (?). I also don't know why he uses stratified sampling since it's a balanced dataset. But more problematic, randomly sampling this across time-periods isn't good practice (and may also have a leakage problem). Lastly, I maintain the blog author has the fundamentally wrong approach on how to work on a data science problem. Always start with the simplest (decent solution) first and and slowly increment from that. Setup proper cross-validation/backtesting approaches on simple models. His initial solution is way overcomplicated and skips the fundamentals. It's like someone going to the gym and lifting the biggest dumbbells' without first having understood the proper technique.