How I Acted Like a Pundit and Screwed Up on Donald Trump

Scott Adams (of Dilbert fame) called it months ago[1]

[1] http://blog.dilbert.com/post/131749156346/the-case-for-a-tru...

Maxim 20 from the Art of wordly wisdom (A must read book)

"The rarest individuals depend on their age. It is not every one that finds the age he deserves, and even when he finds it he does not always know how to utilise it. Some men have been worthy of a better century, for every species of good does not always triumph. Things have their period; even excellences are subject to fashion. The sage has one advantage: he is immortal. If this is not his century many others will be."

This is the age of of the Troll and the response they elicit, Trump read it and understood it, ahead of most and definately ahead of his competitors. He understands the age and is taking full advantage of it. Some saw it coming, and others did not, many who even called themselves experts.

There are no good data sources for politics now. Gallup admits that well over 90% of the people they call for polls won't talk to them. They can't call cell phones for polls. They can try focus groups, but that gets you people who have time to waste.

"we were basically acting like pundits"

So, basically a giant Bayesian prior fail.

edit: got to the part where nate talks about the bayesian failure in more detail. It's good.

> These fundamentals usually consist of economic indicators and various measures of incumbency.

I looked at a couple links, and couldn't quickly find the exact "fundamentals" he refers to. For instance, do they include campaign financing (corrected by peculiar things lke Trump's free media coverage, and perhaps whatever unpaid labor of Sanders' supporters)?

Given that politicians spend enormous time on financing, this is an obvious indicator. But it may be politically incorrect to use models that imply money drives the US.

(Chomsky mentioned, "At the moment, the polls indicate an even race. Usually, US elections can be predicted pretty well by the level of funding, overwhelmingly from the very wealthy and corporations. In the early stages, Bush was far in the lead, not a surprise in the light of the enormous gifts his administration has lavished on a very small wealthy minority and on corporate power. However, that funding gap has reduced considerably in the past months, apparently reflecting concern among elite sectors over the extraordinary incompetence of the Bush planners and the harm they are doing to core elite interests." https://chomsky.info/20041011/)

Here's why this post-mortem is so important:

  The distinguishing feature of the scientific method is not that
  it always gets the answer right, but that it fails forward by
  learning from its mistakes...

  I want to think through his nomination while trying to avoid
  the seduction of hindsight bias.

Kudos to Nate Silver for meticulously examining how he got it wrong, and publicly showing his work so we can all get better.

>"...our early estimates of Trump’s chances weren’t based on a statistical model. Instead, they were what we “subjective odds” — which is to say, educated guesses"

People of al stripes do this in all kinds of situations. We have a hypothesis and find data to support that.

In this case both Trump and Bernie were hard to gauge. Trump just had no prior history (I said to myself) and Bernie was never notable in national politics outside a small core constituency so the default was to discount both because on both sides they were more to the extremes (left and right ends) of their parties.

Also, both Bernie and Trump are the animus the personification of the feelings of large swaths of the electorate --so it goes beyond what they represent as isolated individuals ("movement").

most of these analysts will find it hard to discount their "instinct" even if they want to be data-driven (in a sphere where data can only reveal so much.

Worth noting that pundit Carl Diggler has not had any trouble predicting results even without a fancy statistical model

Trump's (presumptive) nomination was a big surprise to me. However, in retrospect, it looks less surprising. Trump was always going to attract a sizable minority, and there were really no other compelling candidates in the race. While I'm sure that the majority of Republican voters would prefer "not Trump," who that "not Trump" candidate could be was never clear. At the end, well, I am not certain that I wouldn't pick Trump over Cruz, for example.

I don't think 538 gets it even now.

Trump is a celebrity who gets free media coverage and people aspire to. So of course he was a serious contender from Day One.

The fact that he is not polished for the media is a plus to most people.

As one woman said to a news reporter earlier in the campaign, "Well, I could vote for the other candidate. But he's a politician."

I've watched Trump years ago on The Apprentice. What he brought to politics was business sense and marketing know-how, which has been surprisingly absent from political campaigns before.