I said in a previous post here that “turnout matters.” It’s a relevant variable and somewhat obvious. You can’t ignore assumptions. Furthermore, the polls show first and second derivative momentum and the latest ones were still stale. It’s not factored in! The point about correlation, I agree with.

“From a probabilistic standpoint, neither a p-value of .05 nor a ‘power’ at .9 appear to make the slightest sense.”

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2834266

While this should be checked more carefully overall, I think that Silver’s forecast was reasonable.

He gave close to 30% probability for a Trump victory, emphasized the substantial chance for a split between popular votes and electoral votes. Also the error-states is (this time) consistent with

the expected number of errors based on Silver’s probabilities for individual states.

We witnessed some systematic error in polls vs reality and a very strong positive dependency of errors between state. This is something consistent with Silver’s modeling. Probably people gave too strong interpretations to Silver’s 2008/2012 successes (especially the 50/50 success.) Of course, Silver aggregate information from polls (in a way which evaluate the quality of individual polls and estimate the poll’s bias), and the question if improving the quality of individual polls is at all possible and how is a separate interesting issue.

]]>Donald Trump in Back to the Future, Casino, Virtuosity & Simpsons

In case of copyright, see here: https://www.dropbox.com/s/3mn8yhupz692pz9/Final%20Trump%202.mp4?dl=0

Just a few points to temper my attitude toward impossibility:

The major questions in geometry were how to square a circle, trisect an angle, inscribe regular polygons and inverse trig. They can ALL be done with Archimedes’ Spiral, constructed only with a piece of string unwrapped from a circle. Gauss and others somehow thought a piece of string was not a basic instrument of construction, although people used them to make an ellipse since antiquity.

Galois had us believe we could never find the roots to quintic polynomials and Poincare said there was no analytical solution to the three-body problem. All kinds of things are possible with EXISTING computers when previous times thought they weren’t. Numerical methods, optimization algos, non-parametric statistics, perturbation theory, etc… solve just about everything to arbitrary accuracy. Even chiral fermions are being simulated with lattice gauge theory (see Wen). Fairly simple 4-D gauge theory can basically unify physics. We don’t need a supercomputers.

We have the moronic theorems of Cantor, Godel & Turing about impossibility but they are nothing of the sort. Asking a system you don’t know is consistent, if it’s consistent, is moronically stupid. It’s obvious without a proof that logical explosion could have it tell you it is. It in no way says that a formal system could not produce all arithmetical truths. Did you stop beating your wife? Transfinite numbers, “uncomputable” numbers and completed infinities are pure non-sense involving impredicativity mentioned by Poincare and more recently Solomon Feferman. Not a single real number has been truly demonstrated and ultrafinite math rids ALL, and I mean ALL, the paradoxes of analysis and set theory.

P vs. NP is also hardly obvious when you see that the question is poorly posed. “Finite” length programs with pre-computation (a finite set of hard cases) and constant-time jumps (my additional observation about the poor modeling by clumsy Turing machines) even make Don Knuth wonder.

People like Seth Lloyd cannot even understand the meaning of Maxwell’s Demon and thermodynamics. No absolute proofs have been given to refute reversibility. Read Maxwell’s letters and you will understand that he was saying that such a demon could do no work and was not a physical thing but was only entertained to explain a circular reasoning of statistics.

Some things do seem impossible in physics but some of the most prominent examples are not demonstrated by the “skeptic” community. Anyone with a fifth-grade education can understand that energy requirements are too great for deep-space travel, even before considering the relativistic rocket equation with optimistic exhaust velocity. The non-local nature of faster-than-light propagation also seems unlikely when considering how people can’t even understand Gell-Mann’s invocation of Bertleman’s Socks. The world of spookiness is macro and not micro.

]]>