Beyond Bets

Before you begin

Install PySpark.

Mac

To install PySpark on most linux/mac machines just run

pip install pyspark

Windows

PySpark requires java. On Windows we recommend using OpenJDK which can be downloaded from here.

If you are using windows you will also likely need to set the PYSPARK_PYTHON environment variable. When using from within a virtual environment you can set it as PYSPARK_PYTHON=python. Otherwise, you will need to provide the path to the executable.

Install the package

You will need to install the beyond_bets package to complete this project. There are multiple ways to do this, for an editable install navigate to the root directory of the package and run the following:

pip install -e .

To test the install run hello_world.py located in the examples directory. You should see a table with the numbers 1-10 in the console.

Problem

Beyond Bets is a new venture from BeyondTrust to track metrics for sports betting markets. The current codebase works, but the rapid increase in feature requests has left the code in significant need of refactoring. WITHOUT breaking any backwards compatability please add the following new features. It is highly recommended that you improve the codebase as you implement your features.

• The PlayerDaily and MarketHourly transforms have been very popular. Analysts now are requesting PlayerHourly, MarketDaily, and PlayerMarketDaily transforms.

• Management wants to start a new VIP program for our biggest bettors. Implement TopPlayers, which will return the top 1% of all players by total spend in the past week.

• Finance wants to know if the market conditions are changing drastically. Implement BetGrader, which will grade each bet relative to all bets within the same market in the last 15 minutes. For example, if a bet for $15 was placed in market A at 1:15pm, and the average bet size in market A between 1:00 and 1:15 was $10 then the grade for this bet would be $15/$10.

Remember the goal is NOT to perfect the codebase. The goal is to implement the requested features while making some improvements.

Please note, in a real data warehouse the data would sit in physical tables. In order to not write a bunch of junk to your hard drive the datasets are randomly generated.

Lastly, all logic needs to be contained within the beyond_bets package itself. Any code in the examples directory is just to demonstrate how the code might be used.