ฝัง
- เผยแพร่เมื่อ 5 มิ.ย. 2024
- We'll predict the winners of basketball games in the NBA using python. We'll start by reading in box score data that we scraped in the last video. If you didn't watch the last video, you can still download the file (link below) and follow along.
We'll do feature selection to identify good predictors, and train a machine learning model to make predictions. We'll end by computing rolling predictors and improving the model. We'll discuss how you can keep improving the model and predict future games.
Links
Full code and description of the project - github.com/dataquestio/projec...
Dataset if you missed the previous video - drive.google.com/uc?export=do...
Previous video where we did web scraping - • Web Scraping NBA Games...
Chapters
00:00 Introduction
01:00 Reading in box score data
06:10 Preparing data for machine learning
16:10 Selecting the best features for the model
25:31 Creating a baseline model
36:06 Improving performance with rolling averages
41:54 Add in opponent information
51:11 Train a more accurate model
55:08 Improving the model and making future predictions
---------------------------------
Join 1M+ Dataquest learners today!
Master data skills and change your life.
Sign up for free: bit.ly/3O8MDef
Really enjoyed this! I'd love to see another video on how to predict future games. Thank you for the tutorial
I would also love to see actually predicting future games. Thanks for the content!
Wow. Would love to see one on predicting future games. Great video.
Would love to see one on predicting future games. Great video. Very well done
GREAT Video! I will be coding this and Implementing in my personal work. I would love to see a future video on how you go about predicting future games. I would also love to see something just like this for player performance, at a game by game level.
another great vid, would love one on future predictions
Really great video. I learned some good ways to use list comprehensions in pandas to help with column names on top of the scikit learn fits. Thanks for this.
I would love to see the future games please, I enjoy these videos and it helps me learn
Same for future predictors!!! THX FOR THIS!
Great walk through. Would love to see how you update the next values for home and away teams
It would be good if you can put a more detailed guide on Dataquest, to include predicting future matches using rolling averages etc. Would happily sign up just for that!
Awesome explanation, I used this for my class, thank you
Completed the first video, super awesome thank you!!! Does this video help with grabbing player stats and using AVG Reb, PTS, AST, etc to predict stats VS opponents ?
Your videos are brilliant! Horse racing would be an interesting project, using machine learning to predict which horse should win based on the stats for each runner in the race?
Great video. Please make one about predicting future games.
Great tutorial 👌🏾 by any chance did you make the video on how to up date the model
Amazing channel mate! Are you able to demo. how to deploy ML models into production and what we could use to fully automate this end to end? Preferably with systems/platforms that are free to use.
Excellent video that shows you how to use machine learning to identify the correlated factors that determine the outcome using previous games, but is a little misleading because it doesn't actually show you how to predict outcomes of future games. Would love to know where I can find this information, even if I have to pay for it.
Great video, would be great if you could do one but that predicts total points scored, not necessarily in basketball.
simply awesome, thank you
Please do a predictive video for future games 🙏🏻
This was a great video but I would be happy if you would do one for the prediction of future games
When you doing the one to predict the future games i.e. value of 2 in the target column? Thanks
Please do one for future games!!
here I'm waiting the video on predicting future games... maybe someday will come
Hi. I have two questions.
a) Where did you find the data to use for your test?
b) How easy is it for someone who don't know programming to learn python?
Please make a video on how to predict future games!!🙏🙏
Hey! Great video (complicated too, gotta watch second time) I would personally benefit very much from a video on how to use this for future matches, pleaseee!
Great Tutorial!!!!
Hi there, just curious what would you say are the main things to look for when predicting games
Hey,
How long will it take to run SequentialFeatureSelector with the same parameters, but using RandomForestSelector or XGBoost as a model? Couple of hours , days ?
Hi, how could I attach the season to the predictions to see how well the model did for each individual season?
Hi, I'm curious as to why this only results in a 64% accuracy.
For example, something as simple as comparing the records of the teams at the time they've played and predicting the one with higher win% to win would result in around a 68% accuracy for the 2021-22 season.
Is this due to ridge classification?
Hello, what is the algorithm used by the model and where could I get information on the logic behind the algorithm used by the model?? Thank you.
I've been working with this code for about three weeks now and I have successfully scraped all of the player stats too and want to somehow add a 'lineup' feature that looks at the MP of each player and how productive they tend to be to further improve the model. Any chance you would be willing to help me with that?
please make video on how to predict future games
'Customer segmentation in retail using machine learning' please make a video on this topic using real dataset.😥😥🙏🙏
stay strong, Coulibaly is going to be a star
When computing the rolling averages, why did you not use the 'left = X' paramater, like you did in your football predictor video? Don't your rolling averages include knowledge of the current game you are predicting?
How would I filter out rows of games that were in the playoffs so I just have regular season games in the dataframe?
would love to see total score predictor sir.
Hi!
Why did we use ridge classification?
at 38:48 i get an error when running the function find team averages for last 10 games, that i can not resolve. would it have something to do with the error showing in the video, the futurewarning?
I need some clarification here... The data in the training and test set contain the points scored by each teams, how can the model not predict exactly if the game is a win or a loss? It literally just has to check if the team has more points and return true if it does... I am confused
We're predicting the winner of the next game. The algorithm doesn't know what happened in the next game when it is making predictions.
@@Dataquestio facepalm
give this man move views so we get another part !!!!
Why do we need player stats which have max in front of them? What is the purpose of max stats ? Can anyone help clarify please?
How did you decide to chose ridge classifier?
I wonder why using ridge classifier but not logist regression in the SequentialFeatureSelector
hello Please can you show how can we select 2 teams and than the AI would chose who wins , like please write the code in the reply
Doesnt rolling 10 include current game for rolling average? wouldnt that be a leakage?
No, because you predict always the next game
predict future nfl games please
Hi sir,
Where to deploy this project.
Excellent video for learning, but doesn't actually show you how you can predict future games. Future games do not have all the box score stats, which make it difficult project outcomes for the future based upon what this video is demonstrating. Some help or an additional video would be much appreciated showing how to actually use this to predict future games (or games that have not yet occurred).
which ide u r using
JupyterLab
hello brother, can you help me with this line that is generating the following error:
line:
df_rolling = df[list(selected_columns) + ["won", "team", "season"]]
def find_team_averages(team):
rolling = team.rolling(10).media().
return rolling
df_rolling = df_rolling.groupby(["team", "season"], group_keys=False).apply(find_team_averages)
error:
DataError: Cannot aggregate non-numeric type: object.
Did you manage to solve that?
I believe you need to change that line to rolling = team[selected_columns].rolling(10).mean()
this works@@bena.9440
@@bena.9440 Yes it is
INSTANT FOLLOW!!!!
We would really like to see a video on predicting future games in the NBA. Even though this would be a horrible use of PyTorch, I would like to see it done with PyTorch, as well as a wide variety of other machine learning models & technologies (sklearn, etc.). It would also be nice to see some work with regards to this done on Kaggle as well, for example using NBA datasets as well as NCAA datasets.
does this for loop need to be updated for my pc?
for url in standings_pages:
save_path = os.path.join(STANDINGS_DIR, url.split("/")[-1])
if os.path.exists(save_path):
continue
Is anyone facing 'Cannot aggregate non-numreic type:object' error while trying to this:
df_rolling = df_rolling.groupby(["team", "season"], group_keys=False).apply(find_team_averages)
I have the same problem with that error.
i have solved this issue
How did u solve that?
@@akashgahlaut4078
the full
del in begining shuld be like this
del df['index_opp']
del df['mp.1']
del df['mp_opp.1']
del df['mp_max_opp.1']
del df['mp_max.1']
Unrelated, but the previous nba score scraper took like 3 days to scrape 2016-2022. OH MY DAYS.
Yeah, it has to scrape a lot of records (8500), and there is a time.sleep in the loop. Each record should take about 6 seconds to download. There's also a small chance that it will time out after 30 seconds of trying and need to retry. We can guesstimate the runtime with (8500 * 6 + 8500 * .05 * 60) / 3600 = 21.25, so it should take about 21 hours to run.
You could try reducing the sleep time and timeout times for playwright, but there is a risk of getting banned by the server.
How do I add to the date selection line this new season from October to now