Build an NBA Fantasy Projection Model in Python with Pandas

• Jupyter Notebooks overview

• Installing & launching Jupyter

• Creating a New Notebook

• Code & markdown cell types

• Selecting the appropriate kernel

• Restarting the kernel

• Command mode vs edit mode

• Two ways to run code

• Order of execution within the notebook

Python Building Blocks

• Importing modules

• Printing

• Variables & Raw Inputs

• Lists

• Dictionaries

• For Loops

• If, Else Statements

• Functions

• Arrays

• How to import modules

• Giving modules an alias

• Calling specific values from a module

• Using the print statement to output variables

• The value in printing

• Assigning values to a variable

• Changing values in a variable

• Naming variables

• Case sensitivity with variables

• Definition and examples of python lists

• How to write a list in python

• Data types within a list

• Accessing items in a list

• Appending to lists

• Slicing lists

• Definition and examples of python dictionaries

• Looking up values from keys

• Iterating over a list and dictionary

• Printing values with a for loop

• Utilizing break statements

• Using the continue statement

• Pairing if statements with logical conditions from mathematics

• Understanding indentation

• And / Or operators

• Examples of various if, else statements

• How to create a python function

• Creating a points per game function

• Arrays as a special variable

• The benefits of using arrays

• Applying mathematical statements to arrays

If you are coding in python and you come across some type of data, in our case NBA stats, you would be hard pressed to find a better alternative for wrangling, munging and analyzing the data than with pandas. The best part is that it’s open source and free to use.

Essentially, Pandas takes data (like a CSV file or SQL database query output) and creates Python objects with rows and columns (called a dataframe) that looks very similar to a table you’d see in excel. It’s easy to work with and has a lot of methods baked in that make it super useful.

• The two primary types of data structures provided by pandas

• Various inputs that can be used to create a dataframe

• Various file types you can read in to create a dataframe

• How series are different than dataframes

• Creating an empty dataframe

• Creating a dataframe with a dictionary & list

• Using the read_csv() method

• Reading in csv files from a different folder

• Loading a csv with specific columns and indexes

• Using the usecols parameter

• Understanding attributes and methods

• Running through typical attributes (shape, dtypes)

• Head, tail and sample methods

• Using info() and describe() as summary statistics

• Built in pandas dataframe methods

• Calling the sum() method on a single column

• How to select single columns and multiple columns

• Listing columns with list()

• Selecting a single column with dot notation and square brackets

• Selecting multiple columns with a list

• Selecting columns with .iloc and .loc

• Using = syntax to assign values to a column that doesn't exist

• The insert() method

• Deleting columns with del

• Deleting columns with drop() method

• Selecting a subset of columns

• Renaming columns with a list

• Setting new column names with df.columns

• Using the rename() method

• Selecting rows by passing the index location and .iloc

• Using square brackets

• Selecting multiple rows using the slice operator

• Selecting rows using .loc

• The .ix method

• How to remove null values or rows with missing data

• Dropping rows with .dropna()

• The .dropna() parameters

• Adding rows with the .append() method

• What exactly is the inplace parameter and why you should care

• Sorting dataframes based on index with .sort_index()

• Sorting dataframe by a specific column using .sort_values()

• Ascending and descending parameters

• Sorting by multiple column values

• How to filter dataframes based on specific conditions (and / or)

• Using lists to make code more readable

• Examples filtering dataframes for specific teams and stats

• Overview of the .groupby() method

• Grouping by season

• Inspecting the groupby object

• Example getting the mean age for each team

• The split-apply-combine framework

• Finding the sum of a specific column of a groupby object

• Using the .describe() method on a groupby object

• Concatenating dataframes from different sources

• Using the .concat() method with square brackets

• The ordering difference for .concat()

• Difference between concat and append

• Using the ignore_index() method on append

• How to join two dataframes based on common columns

• The .merge() method

• The "how" parameter and various merge types

• Calling the merge directly on a dataframe

• Itertuples() and iterrows() as dataframe for loops

• Returning the data as a series

• How to print a dataframe within a for loop

• Appending values from a dataframe to a list

• How to use the .apply() method

• Applying to an entire dataframe

• Applying to a single column within a dataframe

• Applying to a groupby object

• The axis parameter within apply()

• Exploring the numpy array

• Getting the shape of arrays

• Applying mathematical operations to arrays

• Inspiration from KNN model

• Why we need to normalize player data across seasons

• Breaking down the steps in our model

• Reading in our player data CSV

• Dropping rows with missing values

• Plotting games played histogram with matplotlib

• Function to normalize data

• For loop to apply normalized data function in a groupby object

• Function to calculate euclidean distance

• Testing the function on three NBA players

• Function to find player row with two inputs

• For loop over dataframe to save data into new variable

• Converting a dataframe to a numpy array

• Finding the difference between two arrays

• Vectorizing a function

• For loop to compare a player to multiple players using function

• Dropping duplicates after merging dataframes

• Accessing values in a season list

• Getting the next value in a list based on an input and index

• Appending values to an empty dictionary in a for loop

• Sorting a dataframe based on percent error

• Looping over ten rows of a dataframe

• How to use the getattr() method

• Embedding multiple for loops for various players and seasons

• Creating a function to find a specific player

• Adding a break statement

• Adding in weighted variables to a given function

• Adding a dataframe as an input variable

• Bringing together all the building blocks into a final function

• Looping over ten players to test the projection model

• Creating a dataframe from a dictionary

• Saving a dataframe out as a CSV file

• How to manually calculate the root mean squared error

• Calculating RMSE using scikit-learn

• Using real stats to measure the effectiveness of the model

• Reading in and munging data from a competitor

• Updating dataframe to have identical columns

• Converting a column datatype using .astype()

• Creating new columns with current columns

• Getting the RMSE for competitor projections

• Modify projection model data to match competitor projections

• Running RMSE on both models using same set of players

• Merging dataframe to get player names

• Which variables to modify for new insights

• Function to calculate fantasy points for each player

• Applying function to entire dataframe

• Sorting dataframe based on projection fantasy points

• Saving out dataframe to CSV file

• Overview of value based drafting

• Theoretical example of why VBD is effective

• Function to get our baseline number

• Saving out dataframe as a CSV file for Google Sheets import

• VLOOKUP in google sheets to calculate value over replacement player

• Sorting based on value based points

• Known blind spots for our current model

• Conclusion and next steps