Horse Race Reporting: Comin’ round the bend…
Famed statistics guru Nate Silver and his website FiveThirtyEight are on the “chopping block,” Confider has learned, as ABC News looks to cut costs.
A decision on the future of the famed politics, economics, and sports analysis website is set to be made by the summer when Silver’s contract is up, multiple people with knowledge of the situation told us.
Nate Silver might be great with numbers, but he’s not a great pundit. The trouble is that people are unpredictable and while after the fact Nate can say with confidence what happened, his predictions of what will happen are uniformly bad. Nate really doesn’t understand actual humans, and I believe he’s even said that.
The oft-combative Silver, who has come under renewed scrutiny for his site’s polling selections, now faces an uncertain future as news division boss Kim Godwin reviews the publication, which has never turned a profit.
That last bit startled me, so… when ESPN sold 538 to ABC news in 2018, the W$J reported that “[potential] buyers have been told it is losing about $6 million annually on revenue of about $3 million.”
Mock Paper Scissors 
Weirdly, his ‘one weird trick’ (baseball statistics) turned out to be not so useful outside of baseball.
It is stunning how people reach a certain amount of fame and are allowed to fail spectacularly over and over again.
There Is A Club, and I’m most DEFINITELY not in it.
LikeLiked by 1 person
His methodology for aggregating polls was innovative at the time, but the flaw with it became apparent quickly. If the polls were bad, the conclusions were bad. And as Wingnuttia learned to lie to the polls (and as everyone —started using mobile phones nearly exclusively), the samples were not really random. This last round, Silver included partisan polls commissioned by the campaigns, which really are vanity projects telling the candidates what they wanted to hear. They guys at E.V. stripped campaign polls out of their aggregates and did much better than Nate.
The entire polling industry is in free fall, and I think Nate owns some responsibility for that. Everyone turned to him as an oracle and he did not effectively dissuade them.
Rgds,
TG
LikeLiked by 1 person
The basic rationale of polling meta-analysis is correct; this is the same thing as combining a bunch of different clinical studies into a larger meta-analysis. This is a perfectly valid methodology to increase your sample size…so long as you don’t include a bunch of garbage studies.
(this is how the reality-based medical professions rapidly figured out that super-hydroxy boner pills and horse paste were NOT, in fact, sovereign cures for COVID, once they excluded the fraudulent, poorly designed or inaccurate clinical studies.)
You’re dead on, though; I think Nate included the partisan ones based on his naive belief that a campaign would WANT to do accurate polling to suss out their weaknesses and strengths, when what they WERE doing were partisan push polls designed to fool Nate Silver so as to drive the overall conventional wisdom and reporting.
Also, he’s a slave to the past: pretty much always in modern history, the incumbent president has lost the House by large numbers, so you can’t go wrong predicting that outcome. But there’s that little disclaimer they put on investment funds
“Past performance is not an indicator of future performance.”
LikeLiked by 1 person