Nate Silver is my new hero

[maddog42]

I picked up Nate Silvers 2002 book The Signal and the Noise today. Read about 40 pages. This is on page 11:

"A long-term study by Philip E. Tetlock of the University of Pennsylvania found that when political scientists claimed that a political outcome had absolutely no chance of occurring it nevertheless happened about 15 percent of the time."

He also went on to say that poly-sci guys are probably still better than TV pundits. Note the date of the book, after Gore lost but before Trump.

[davew]

Nate had Hillary at 90%

[maddog42]

Nate had Hillary at 90%

I don't think so. He gave Trump a 29% chance about 48 hours before the election. I have stated this percentage in other posts.

[maddog42]

https://fivethirtyeight.com/features/election-update-dont-ignore-the-polls-clinton-leads-but-its-a-close-race/

[maddog42]

Handicappers take note, chapter 8 of this book is about Bayesian probability.

[_______]

Definitely a worthy read. If you enjoy his chapter on Bayesian probability, you should try beating 538’s pro football algo next season. I hung in for a little while before blowing up completely after some overconfidence in games where key players were injured and the team won anyway. Their algo doesn’t take injuries into account at all. I overweighted their importance and lost some games I rated at 98% probability. Don’t do that.

I think you hit a zero by accident. The book was published in 2012.

[davew]

I don't think so. He gave Trump a 29% chance about 48 hours before the election. I have stated this percentage in other posts.

I was wrong - 88.1% on Oct 17

https://projects.fivethirtyeight.com/2016-election-forecast/

[Robert Fischer]

Nate Silver is quality. He actually has insight.

A percentage of the 'pundits' we see are average or below average in terms of insight, even though they are authorities at these fields.

An authoritative position and maybe a little charisma is generally sufficient to sell. Insight such as Silver's is often considered 'superfluous'.

[maddog42]

Definitely a worthy read. If you enjoy his chapter on Bayesian probability, you should try beating 538’s pro football algo next season. I hung in for a little while before blowing up completely after some overconfidence in games where key players were injured and the team won anyway. Their algo doesn’t take injuries into account at all. I overweighted their importance and lost some games I rated at 98% probability. Don’t do that.

I think you hit a zero by accident. The book was published in 2012.

I definitely goofed on the publication date, but not on his chances of Trump winning. Thanks for the correction and the worthy insights. Is their algorithm published or just described?

[maddog42]

Nate Silver is quality. He actually has insight.

A percentage of the 'pundits' we see are average or below average in terms of insight, even though they are authorities at these fields.

An authoritative position and maybe a little charisma is generally sufficient to sell. Insight such as Silver's is often considered 'superfluous'.

I am impressed with his predictions and his reasoning. He makes a big deal of predictive power and being brutally honest about evaluations.