How Azure ML Helped Predict the Results of the NCAA "March Madness" Tournament

Play How Azure ML Helped Predict the Results of the NCAA "March Madness" Tournament


What if you could improve your “March Madness” bracket by using Azure Machine Learning? Tune in for this insightful interview as Jennifer Marsman welcomes Damon Hachmeister to the show as they discuss just how Damon used Azure ML to improve his rankings in last year’s NCAA tournament.

Think you can do better this year? Check out Azure Machine Learning now!

  • [0:56] Tell us a little bit about the project. How did this experiment get started?
  • [3:04] DEMO: How to build a bracket using Azure ML

 Follow the conversation @msdev
 Become a Fan @
Generic Episode Image Subscribe to our podcast via iTunes, Windows Phone Marketplace or RSS

If you're interested in learning more about the products or solutions discussed in this episode, click on any of the below links for free, in-depth information:

Websites & Blogs:




The Discussion

  • User profile image

    Is that experiment publicly available to see?
    Very interesting indeed.

  • User profile image
    David Blundell

    When I click on an email the first thin I want to know is what it is?
    Azure ML. NCAA. March Madness. Just gobbly gook. What about two lines of explanation of what it is that you are offering. You are not the only ones who live in a small world. I do hope that I am not missing something good?!!

  • User profile image

    hola, muy interesantes estos avances, gracias

  • User profile image

    @Francesco: Unfortunately, the experiment Damon walk through is not published online.

    @David Blundell: My apologies.  AzureML is Azure Machine Learning, a tool that allows you to perform enterprise-quality machine learning from a browser.  The documentation is at  "March Madness" is a college basketball tournament that happens every year in March in the United States.  It is known for having great upsets (a low-ranked team beating a high-ranked team) so it's popular to create a bracket to predict who will win and proceed to the next level of the tournament.  More information at

    @eugenio: Gracias!

  • User profile image
    Byron Rogers

    It looked to me that the accuracy of 0.7 was on the training data but on the holdout testing data it was 0.57, slightly better than a coin flip. Or did I miss something?

Conversation locked

This conversation has been locked by the site admins. No new comments can be made.