
Last night
as the primary results began to show Bernie outpacing Hillary Clinton in
Michigan I said to m'self: WTF? Didn't Nate Silver predict that
Sanders had, essentially, no chance on this one?
Yeah, that's exactly what he predicted. When I plugged
into Sliver's website, fivethirtyeight.com, to verify what I thought he said,
here's what I found: according to Silver, Hillary had a greater than 99% chance
of winning Michigan; Bernie less than 1%.
This is the way Silver was calling the Mich. Democratic primary as of
at least
9:30pm PST, Mar082016
Now, to be brutally honest, in these primaries that are down
to two candidates and where the delegates are split according to precinct, it's
not all that hard to come damn close to making a close prediction. A monkey with
a dart board could do a fair job most of the time and have some absolutely
stellar successes occasionally. And some equally stellar failures. The
problem is that nobody says what it means to "win." Does it mean
having the most votes? The most delegates? The most precincts?
With so many ways to define "win" you'd have to be basketcasebrainded
not to be able to make a pretty close prediction in one category or the
other. In fact, as of 10am PST on Mar09 as I write this (14 delegates
still not counted), if you choose to use delegates as your criterion, Hil has
actually won Michigan: she has 68 delegates to Bernie's 65. But she has
48.4% of the votes to Bernie's 49.8%, which, apparently, according to Silver and
the MSM means she lost Michigan. But hold the phone, Barney  percentage
of the votes don't mean squat when it comes to the nomination. I'd much
prefer to "lose" with a smaller total vote and go to the convention
with more delegates.
From HuffPo as of 10:25am PST, Mar0916
Note the fine print: Silver was not predicting that Hil would
get 99% of the delegates or votes or anything, he was predicting that it was
more than 99% certain that she would win. If she had gotten just one vote more
than Bernie, Silver would be claiming success. What bullshit this is.
Let's go back to the monkey. If you had a dart board
divided into 20 pies randomly numbered in 5% increments from 5% to 100%, that
monkey would have a  pulls out calculator, tap, tap, tap  95% chance of
making a closer guess on the Michigan contest than Nate Silver did. And
note that Silver's 99% wincall for Hillary was 50% worse than the worse
guess of the polls he relies on (Mitchell Research & Communications, Mar06
called Hillary by 66%.). IOW Silver's much ballyhooed algorithm for
weighting the poll numbers is actually a statisticalbullshit multiplier,
otherwise known as a SBSM in all but the most effete of pedantic statistical
circles.
Silver  much like Rubio, Bush, Carson, and Christie  is
losing it big time this election. In January this year Leon Neyfakh
of Slate
called out Silver for being wrong, wrong, and wrong with respect to Donald
Trump. It's a case against Silver that is worth the read, certainly if you
are using Silver's BS to place bets in these elections, which, just for the
record, I am not.
In September last year, Silver
said Trump had a 5% chance of getting the GOP nomination. Well, OK
. . . duh . . . at that time there were 17 frikin' Republican candidates,
meaning just by chance each one of them had roughly a 5% chance of getting the
nomination. Monkey, dart board again. On Nov23 there were 14
candidates left in the race, which would mean each had, on average, a 7% shot,
but the polls showed that Trump was running about 4x that at 2530%. And
yet Silver famously
said that by his analysis a 20% chance for Trump was "substantially
too high." This was not an analysis based on math or crunching numbers, it
was just his own biased opinion as a proClinton guy. In one of the most
gormless arguments imaginable, Sliver said that 30% of Republicans polling for
Trump was about the same proportion  68%  of Americans who think the
Apollo moon walks were faked. That sort of flap suggests to me the guy is
a huckster.
Speaking of hucksters, Silver is this generation's Jeanne
Dixon, even if this generation doesn't know who she was. Dixon was a
wildly popular soothsayer. A spoofer. She was one of those people
who make thousands of predictions and eventually, just by sheer, unmitigated
chance, gets a big one right. After that all of the suckers come
running. To borrow another monkey metaphor  infinite monkeys with
infinite typewriters given an infinite amount of time will write all of the
works of Shakespeare. If you chatter on for long enough, eventually you'll
say something brilliant, and that's what people will remember because your PR
team keeps reminding them, and that's what they'll judge you by. But the
real measure of any person's prescience  statistician or other type of
soothsayer  is how far off the mark are they when they muff it. When you
assert on the day of the Michigan primary that Sanders has less than a 1% chance
of winning, and he wins, you've just told the world that you don't know what the
hell you are talking about.

