By JD Chang
Using only daily scrapers for news, social, and sentiment, TrendPo correctly predicted 24 out of the 33 Senate races in this year’s elections.
See the full results here.
Looking at the data, when we ran our algorithm on a 14 day input (10/23 to 11/6), we hit on 72% (24 out of 33) of the Senator races on Election Day.
As a frame of reference, Nate Silver’s 538 Blog correctly predicted 31 out of 33 Senate races. Interestingly, the two that Nate Silver got wrong (Montana, North Dakota) we got right. However, we also had some big misses like Ted Cruz (TX) and Roger Wicker (MS). We’ll be diving more into the reasons why in future blog posts.
Overall, we definitely see a correlation between a candidate’s “buzz” and their election results. This is even more apparent when you add in the 100% accuracy rate of predicting Governor races. The advantage here, of course, is that TrendPo Rank can be used to measure a politician’s momentum and buzz on a daily basis…not just every two or four years.
TrendPo runs a daily algorithm that scrapes National Media, Beltway Media, Facebook pages, Twitter profiles, and sentiment analysis to create a daily rank of all national politicians. There is no polling, no sampling, and no bias. It’s a calculation we run every day against every politician and select issues. We will continue to run it post Nov. 6th to track buzz and movement of all winning candidates and current incumbents.
Visit TrendPo to see full, daily rankings.