Data Walrus

February 1, 2018

A shadow of a doubt

It's that time of the year where many wait for a groundhog to pop out and predict the weather for the next six weeks (I wonder if meteorologists get jealous of all his attention?). While this all seems very silly, it can be a real source of optimism for those dealing with frigid temperatures. In the past I've gotten my hopes up that Punxsutawney Phil (the groundhog) will predict an early spring and rid me of having to defrost my car for 20 minutes.

Well, being someone who has to analyze everything, I decided to see how accurate this little woodchuck has been with his prognostications. So, let's get to those insights!!

To accomplish the task of measuring Phil's accuracy I needed some key data points. I first had to find a record of all his predictions. It turns out there are some pretty detailed records of this information over at the Stormfax Almanac. Next, I needed US temperature information for February and March of each year. The fine folks at the National Oceanic and Atmospheric Administration (NOAA) had a really great resource for this.

Before we move further it's important to note some relevant information. When Phil appears he either sees his shadow, therefore predicting six more weeks of winter, or does not, therefore predicting an early spring.

Now that I had all this information I next had to determine the logic for assessing the accuracy of Phil's prediction. To do this I used the NOAA data to see whether February and March were colder or warmer than the national average. This would help me determine if there was an early spring or long winter. The basic logic was if either February or March had temperatures above the national average then it was an early spring. If both February and March were below the national average temperature then it was a long winter.

With all my data and logic established I was now able to visualize the distribution of Phil's predictions and his overall accuracy.

Phil is not really great with his predictions, coming in at 37% accuracy. If you left it purely to chance, say, by flipping a coin, then you would have a significantly better outcome than Phil. Another interesting insight is how infrequently Phil predicts an early spring (13% of his predictions). This does seem to be changing with more early spring predictions in recent years - maybe Phil is more in tune with climate change?

After all is said and done, I'm still going to check the news tomorrow to see if Phil saw his shadow or not. I'm predicting he won't…but that's just a guess.


# Back