[ExI] Nate Silver

[John Clark]

On Mon, Sep 5, 2016 at 10:32 AM, spike <spike66 at att.net> wrote:

> Nate Silver uses only polls.


I like Nate Silver because he's a number crunching sort of guy and because
of all the political pundits he has proven to be by far the most accurate.
In the 2008 Presidential election Silver correctly predicted the outcome in
49 of the 50 states, in 2012 he was 51 out of 51, he got every state right
and the District of Columbia too.  Nobody else came close. And yes he only
uses polls in his "polls only" forecast, but not in his more accurate
"polls plus"  forecast. And Silver doesn't just use one poll but dozens,
and he knows that some polls have more scientific protocols than others so
he gives some more weight than others, he also knows that some have
historically favored the Democrats and some the Republicans and he knows by
how much and makes adjustments. Silver changes the probabilities every day
as a good Bayesian should because new information becomes available every
day.

As of today with "polls only" Silver says Trump has a 28.9% chance of
winning, close to what the betting odds of Trump winning are, 27.6%.
Betting odds have a good record, better than any single poll, but not
better than Silver's "polls plus". In "polls plus" he makes further
adjustments using economic forecasts, endorsements,  short lived
 convention bounces, and early voting which will start in some states in a
week or so. Using "polls plus"  Silver says  Trump will do better than what
"polls only" indicates, Silver says Trump really has a 30.8% chance.

> He has never done a poll in which one candidate faced a looming email
> dump, nor is he dealing with Friday’s FBI report,


The report where the *REPUBLICAN* FBI director said no crime had occurred?

John K Clark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20160905/b0ea23bd/attachment.html>