I’m calling out Nate Silver! That’s right, I’m challenging the reigning king of statistics and predictive modeling himself. What’s my beef with Silver? In truth, I don’t really have any beef with Silver. I have great respect for him and admire what he, his predictions, and his writing have done to increase visibility of the power of data, analytics, and statistics. However, every time I read an article about him or one of his predictions, the author never fails to point out the fact that Silver accurately predicted the winners of 49 of 50 states, plus the District of Columbia, in the 2008 Presidential election, then predicted all 51 contests in 2012. While this seems like an incredible feat only possible for a genius, I have always wondered how difficult it could really be to accurately choose the winner of each state.

So I’m here to challenge Nate Silver. As we head into the final months of the 2016 Presidential election, I am going to try my hand at predicting the winner of each of the 51 contests. My theory (which could very well be wrong) is that it can’t be that difficult. Here are some thoughts on this theory:
• No-Brainers – At least 40 or so of the contests are no-brainers. States like Maryland and Massachusetts have gone to Democrats at least the 6 last elections, while states like Idaho and West Virginia can be counted on to go Republican.
• Swing States – This leaves 11 or so contests that are truly swing states and will, therefore, require a bit more attention: Colorado, Florida, Iowa, Michigan, Nevada, New Hampshire, North Carolina, Ohio, Pennsylvania, Virginia, Wisconsin
• Key Data Points – Silver employs a proprietary statistical model to predict the elections, but I think the use of a just a few key data points is enough to make an accurate forecast (I’ll talk about those data points later).

Now, of course, I could be totally wrong—perhaps it’s even likely that I am—but that's why I’m doing this experiment in the first place. I really want to know if my theory is correct and I figure I’ll have a bit of fun along the way. If I am wrong, I’ll humbly admit defeat and acknowledge that Nate Silver is way smarter than me (okay, I think it’s safe to already say this is true, regardless of the results). If I’m right, well perhaps I’ll write a best-selling book, create a website dedicated to predictive modeling, make the rounds on the 24 hour cable news channels, etc. But, in all seriousness, right or wrong, when it’s all said and done, I will share my findings and analyze my hits and misses and lessons learned.

My Approach
Here are the keys to the approach I’ll be taking:

1.      Data Only – I will only use data to make my predictions—I won’t be listening to Sean Hannity, Rachel Maddow, or any of the pundits; they have other motives for their predictions and don’t pay close enough attention to the data (Nate Silver discusses this in The Signal and the Noise and I completely agree with his indictment of pundits). In addition, I will not let any news item, scandal, etc. impact my predictions.
2.     Data Sets – I will only use two key data sets to make my predictions, most of which I’ll get from either www.realclearpolitics.com or www.270towin.com.
o    Results of Past Presidential Elections
o    Poll Averages Per State
3.      Simple Analysis – There will be no advanced statistical models—just some very simple analysis of the above data sets.
4.     Limited Effort – I will spend very little time and effort on my analysis.
5.      Silver’s Predictions – While I will likely take a look at Silver’s predictions from time to time—just to keep an eye on my competition—I will not allow his predictions to sway mine in any way (you’ll have to trust me on this one).

My First Prediction
After spending some time with the data noted above (probably about an hour), I have completed my first set of predictions. I’m providing my results here in both a list and a geographic map. I used Tableau Public for the map, so feel free to interact with it here: https://public.tableau.com/profile/ken.flerlage#!/vizhome/Election2016_7/Election2016. If you’d like to compare, you can also view Silver’s predictions here: http://projects.fivethirtyeight.com/2016-election-forecast/#plus

Note: Coincidentally, my first set of predictions lines up quite nicely with Silver’s first set of predictions, which he just recently unveiled (Update: Silver updated his projections since this article was first posted, so his no longer match mine exactly).

Clinton - California, Colorado, Connecticut, Delaware, District of Columbia, Florida, Hawaii, Illinois, Iowa, Maine, Maryland, Massachusetts, Michigan, Minnesota, Nevada, New Hampshire, New Jersey, New Mexico, New York, Ohio, Oregon, Pennsylvania, Rhode Island, Vermont, Virginia, Washington, Wisconsin

Trump - Alabama, Alaska, Arizona, Arkansas, Georgia, Idaho, Indiana, Kansas, Kentucky, Louisiana, Mississippi, Missouri, Montana, Nebraska, North Carolina, North Dakota, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, Utah, West Virginia, Wyoming

My intention is to update my predictions around the end of each month, then once more right before the election. So, check back regularly…and wish me luck!!

Ken Flerlage, June 30, 2016
Originally Posted on LinkedIn: