How To Make A Predictive Model In 5 Easy Steps

Step 1: Choose your language

There are lots of programming languages to choose from. For our data modelling workshops we work in R and Python, as they’re both relatively easy to learn and designed for working with data.

If you’re new to these languages, here are some resources that will help get you set up.

Language 1: R

• What is R?
• Download and install R – get the language set up on your computer
• Download and install RStudio – you’ll need a program to develop in, and this one is custom-designed to work with R
• Take a look at the some of the existing R libraries you can use if you want to connect to our API, including abettor and our Data Scientists’ R repo.

Language 2: Python

• What is Python?
• Download and install Anaconda Distribution – this will install Python and a heap of data science packages along with it

Step 2: Find a data source

Finding quality data is crucial to being able to create a successful model. We have lots of historical Exchange data that we’re happy to share, and there are lots of other sources of sports or racing specific data available online, depending on what you’re looking for.

For our workshops we use historical NBA odds data from the Exchange (which you can download directly from here), along with NBA game data from a variety of sources including:

• ESPN.com
• NBA.com
• basketball-reference.com
• Stattleship’s API

Step 3: Learn to program

It’s daunting at first but there are lots of resources out there to help get you started. These are some of our favourites if you want to learn to use R or Python for data modelling:

• Dataquest – free coding resource for learning both Python and R for data science
• Datacamp – another popular free resource to learn both R and Python for data science
• Codeacademy – free online programming courses with community engagement

Step 4: Learn how to model data

We’ve put together some articles to give you an introduction to some of the different approaches you can take to modelling data:

• This Introduction to Tennis Modelling gives a good overview of ranking-based models, regression-based models, and point-based models
• How we used ELO and machine learning as different approaches to modelling the recent World Cup
• We also have resources on our GitHub repo, where our Data Scientists have shared modelling tutorials using AFL and soccer data, along with a R repo for connecting with our API

Step 5: Get your hands dirty

The best way to learn is by doing. Make sure you have a solid foundation knowledge to work from, then get excited, get your hands dirty and see what you can create! Here are a final few thoughts to help you decide where to from here:

• Make sure you’ve got your betting basics and wagering fundamentals knowledge solid
• Learn about the importance of ratings and prices and get inspired by the models created by our Data Scientists
• Take a look at our Automated Betting Station and consider how you could use our API in building and automating your model
• Read about how successful some of our customers have been in their modelling journeys