Nate Silver Considered Harmful

As the 2016 Republican Presidential Primary comes to an end, so has Nate Silver’s credibility. Silver is known as the whiz kid who predicted the 2008 and 2012 general presidential elections using historical data.

As the 2016 Republican Presidential Primary comes to an end, so has Nate Silver’s credibility. Silver is known as the whiz kid who predicted the 2008 and 2012 general presidential elections using historical data. It’s just sad that Nate Silver is consistently wrong lately because of his overt bias (that apparently he can’t recognize).

Luckily someone replied and called Silver out for his bias bullcrap:

Isn’t it ironic how someone who’s opinion based on facts is discredited the second they stop using empirical evidence? Anyhow, Nate is scraping at any and every explanation to save his credibility.

Too bad nobody listens to him anymore, at least after how wrong he was:

Biased Nate Silver

[Mirror Here]


Anyhow, enough ranting, I’ll criticize something that Silver actually does. Take this gem about Trump’s odds of winning the Republican presidential nomination that Silver published March 21, 2016:

To best reflect the group consensus, we used an “olympic average” … in some cases the average reflects a number of delegates that it would be mathematically impossible for Trump to achieve. [Source: FiveThirtyEight]

In context, using an Olympic Average isn’t a bad first start – it’s certainly better than using just a regular average. But using an average in this situation where your data swings between 0 and numbers like 17 isn’t smart at all.

If numbers you’re using are “mathematically impossible” then it certainly isn’t a smart idea to use them in predictions. And given that the outcome (this election) is only happening once, it’s far better to use a deterministic method for determining Mr. Trump’s probability of winning the Republican nomination before the 2016 GOP National Convention. It’s better in some cases to just swing between 0% and 100% instead of use a soft and mushy unattainable number. It’s not like we’re going to have 1000 Year of 2016 Republican Presidential Primaries and to all thousand parallel universe candidate’s delegates then divide by a thousand.

There’s no excuse for Silver’s methods. It’s nearly trivial to create software to simulate elections – so for any serious political estimator or entity which plans on predicting outcomes of elections or deals, there is simply no excuse to build engines to simulate events and calculate probabilities.


Nate, here’s a predictive model: analysis and extrapolation of the past doesn’t work for anomalies of the present. And everybody has known since Day One that Donald Trump is an anomaly. Just like Ronald Reagan was an anomaly.

Nate Silver, your predictive models are irrelevant and useless because this election isn’t ceteris paribus. A key variable has changed: a non-establishment candidate is winning (sorry Ron Paul fans).

For a while I thought FiveThirtyEight would be a shining example of real-world fact-based reporting in the Mainstream Media. Apparently I was wrong to think that FiveThirtyEight would become anything other than just another news outlet spewing out lies and a left-leaning narrative. It’s a disgrace that someone who built credibility from numbers and data had that same credibility destroyed by an overt bias.

By the way here’s my predictive model for this election: All roads lead to Rome.