Trump. Because statistics isn't math

Let me just start this touchy subject by saying I didn't vote in the election, which was the winning choice this election cycle, because I didn't like any candidate. George Carlin says it best. That said, the coverage of the election and more importantly the polls were obviously biased/rigged/wrong, pick the word you like best. And given the absolute shock by soooooo many people, I have to revisit the truth, as I mentioned just last post, that statistics of polling isn't math. That showed that not only do the +/- percent errors not mean what you think, but trained hacks statisticians can have exactly the same data and come up with several different answer. That's not how it's done in math. Previously we saw the BS field of statistics make a major error with respect to Brexit, when virtually everybody got it wrong. Trillions of market cap were wiped off the financial markets. That great job was followed up with the recent US election. Once again statisticians were wrong-- BADLY wrong --with over a trillion in losses for the week. Check out Goldseek's summary of polling data before the election. TheWrap also has plethora of pre-election predictions for electoral college, percentage chance of winning, and percentage to win by: L.A. Times: Clinton 352, Trump 186, Moody’s Analytics: Clinton 332, Trump 206, Rothenberg & Gonzales: Clinton 323, Trump 197, Sabato: Clinton 322, Trump 216, FiveThirtyEight: Clinton 302, Trump 235, Associated Press: Clinton 274, Trump 190, New York Times Clinton 85 percent chance of winning, Hypermind: Clinton 74 percent chance of winning, PredictWise: Clinton 89 percent chance of winning, Reuters/Ipsos States of the Nation: Clinton 90 percent chance of winning, ABC News Political Analyst Matthew Dowd: Clinton 95 percent chance of winning, Bloomberg: Clinton +3, CBS News: Clinton +4, Fox News: Clinton +4,Reuters: Clinton +3, ABC/Washington Post: Clinton +4, Monmouth: Clinton +6, Economist/YouGov: Clinton +4, Rasmussem: Clinton +2, NBC News: Clinton +6.

Those numbers were backed by strident rhetoric regarding the impending, decisive victory of  Clinton. NYPost on the pundits who got it so wrong, including: "Deadspin columnist and GQ correspondent Drew Magary went so far as to publish a scathing piece about the Republican candidate on Sunday — titledDonald Trump Is Going To Get His Ass Kicked On Tuesday.”" TheNewser mentions "Princeton ace Sam Wang, who in mid-October tweeted, "It is totally over. If Trump wins more than 240 electoral votes, I will eat a bug."" . Let's hope he is a man of his word and does it. But don't hold your breath. Even some of the margins of victory were way off: The Telegraph reports "In Iowa, the result was over six percentage points more in favour of Trump than the final polling average. In Ohio - the key bellwether state that has sided with the winning candidate in every presidential election since 1964 - the average polling was five points out....Non-college educated white people make up 41.2 per cent of Indiana's population, for example. The polling gap here was 8.6 percentage points - with the polls giving an average 10.7 point lead to Trump, compared to the actual 19.3 point lead he managed...Wisconsin...was a huge 7.4 percentage point gap between the polls and the result here: the average polling on Real Clear Politics before the vote had Clinton winning by 6.5 percentage points, when in fact Trump claimed the state by 0.9 points."

Which led TheWrap to ask how is it possible to be so wrong? While statisticians are confident of their pre-election prediction, once their wrong, they have no trouble coming up with a excuse reason as to why. From the link "Another theory is that many Trump supporters were simply ashamed to admit that they were going to vote for the GOP nominee and lied to pollsters. Trump is one of the most polarizing figures in American history and many voters wouldn’t even admit it to a complete stranger looking to conduct a poll.". That was a Brexit excuse, too. And yet they were still so sure of their prediction, rather than reduce their percentage of winning to reflect the possible problem. Maybe next time? Some claim the results are lies. Maybe some, but with trillions of dollars of investment capital being lost through Brexit and the US election, there is a lot of "smart money" that should have known better. When you mess things up that badly in math, you look through your work, find the error and fix the problem. In statistics, you find an excuse, or better yet multiple excuses as to why you're wrong and do it again. With each reason you hear ask yourself "then how can they be so sure?".

Of course, some will state "nobody saw it coming". But that's true only if you watch MSM. Check out the Keiser Report, for example where at the 9:35 mark they mention Chris Whalen and Jim Rickards. Chris Whalen is interviewed there. Jim Rickards had a more guarded prediction here. But nobody was more vocal than Andy Hoffman. And, oh yeah, he predicted Brexit, too. But these guy aren't pollsters. Curious, There's a lesson in there somewhere.

And it's not like it was difficult to see some problems: Kaine canceling a Sarasota, Florida rally after only 30 people (about half of which were press) showed up at his West Palm Beach rally. Meanwhile Trump had a different experience: "Republican presidential nominee Donald Trump held a massive rally Tuesday evening at an airplane hangar in Melbourne, Florida to a massive crowd of about 10,000 supporters-while an equal number were turned away when the fire marshal declared the event at capacity....Neal interviewed the last person in line allowed in, Rachel Padrick-Miller. She had been in since 2 p.m. When the fire marshal cut off the line she said ‘thousands were trying to break through the line…jumping the barricades.". And Clinton? "The New York Times features a photo of Hillary Clinton being welcomed to an early voting site in Pompano Beach, Florida on Sunday…
Surrounded by screaming TRUMP supporters!
". Not to mention people chanting "Lock her up!". And see the 11:33 mark of the Keiser Report link above where they witnessed so much more Trump support in North Carolina that they couldn't reconcile with what the pollsters were predicting guessing.  This attendance deficit was in lots of states which witnessed packed Trump events versus a few hundred Clinton supporters. Statistical hacks saw nothing wrong with there models despite the fact that these are the people more likely to go out and vote [See Reuters story below]. They were also aware of Wikileaks and the other umpteen excuses and yet they still were sure of a Clinton win. Now they will "fix" the models. But don't expect the predictions to get any better 10 years from now--statistics isn't math.

Here are some stories that caught my eye:

  • The World Chess Championship has begun. Magnus Carlsen, the champion, takes on Sergey Kajarkin. The official site is here, but I'd recommend Chessbomb for following the games (unless you're willing to pay up for coverage).
  • The best chess is being played between the computers. Komodo has already been dethroned. This year, Stockfish battles Houdini for the 9th TCEC Superfinal. Follow it here.
  • reports "Professors across the nation are cancelling classes Wednesday to help students deal with their emotional distress as they come to grips with the "shocking" election of Donald Trump. ". Poor, poor snowflake.
  • The Atlantic has a more in depth look of the various excuses reasons why pollsters were so wrong.
  • TruePundit on Newseek's election goof: "The partisan hacks at Newsweek are recalling 125,000 copies of its Hillary Clinton “Madam President” issue. The brutal mistake, printing and shipping the issues to wholesalers and retailers weeks before the election, is estimated to cost Newsweek a loss of approximately $500,000. But Newsweek is not new to making lousy management and fiscal decisions. Newsweek claimed it made a Donald Trump edition too although it never shipped it or produced a verified copy of it. Now Newsweek says it will print and ship the Trump edition and hope to break even on the misadventure."
  • ZeroHedge on fudged polls: "The problem, said Cliff Young, president of Ipsos Public Affairs US, the polling partner of Reuters, came down to the models the pollsters used to predict who would vote - the so-called likely voters. The models almost universally miscalculated how turnout was distributed among different demographic groups, Young said. And turnout was lower than expected, a result that generally favors Republican candidates. Ultimately, missing that shift in the state polls tripped up the predictions. It also highlights how the otherwise empirical process of polling rests on a subjective foundation. Each pollster must make a decision about turnout. Their decisions are informed by historical voting patterns. But the actual turnout in each state is unknowable before election day."