Tag Archives: academy awards

Broken models lead to failure

The Sunday keynote interview featured Nate Silver of fivethirtyeight.com, a political website that was created prior to the ‘08 presidential election that used a unique algorithm to predict the outcome. His career started by using an innovative approach to analyze baseball statistics, and this election allowed him to take the approach to a new level. By factoring in the complexities of humans beings, his statistical analysis changed political polling forever.

One of the biggest takeaways occurred during a discussion of his projects since the election. While using his method to predict the outcome of the latest Academy Awards, he realized it didn’t take into account the Academy’s reluctance to give an award to someone who had previously won. In the category of Best Supporting Actress he realized this error would result in the wrong prediction, breaking his first rule: never use a model that predicts something that you know will be wrong. “If you know you’re gonna be wrong, keep working on you F*@#&! Model.”

This statement applies to so much more than just gathering statistics. Whether it’s voter polls, educational curriculums, business models, group projects, personal interactions, etc. if you know in advance that the approach you’re taking will not lead you to the solution you’re trying to achieve DON’T KEEP DOING IT JUST TO DO IT.

Unfortunately, changing this model will likely be difficult: if it was easy you’d already be doing it, but instead it requires going back to the drawing board and reassessing the process. At times this might be scary as hell or hard to embrace, but saving the time, energy and resources that are spent building an unsuccessful finished product will be so much more useful in the end, and will truly allow for a sense of personal or organizational pride.

Advertisements

Leave a comment

Filed under Uncategorized