Wednesday, January 11, 2012

Results for prognostication #2:
New Hampshire primary

There are three things about this week's guess that should be noted.
  1. I got beat by Nate Silver's guess, 88.8% to 88.2%.
  2. I did much better than I did last week.
  3. I got Gingrich over Santorum and Nate Silver went the other way.
Here are the final numbers.

Mitt Romney: 39.3% (Silver beats me by 2.9%)
Ron Paul: 22.9% (I beat Silver by 2.7%)
Jon Huntsman: 16.9% (Silver was at 17.0%, I was at 16.7%, so he was 0.1% closer)
Newt Gingrich: 9.4% (We both overshot, I lose by 1.2%)
Rich Santorum: 9.4% (We both overshot, I win by 1.1%)
Rick Perry 0.7%: (Again, we both went high, I win by 0.1%)
Buddy Roemer: 0.4% (Silver was silent, I went high and lost 0.8%)

I learned several things this week, mainly about thinking like a mathematician instead of thinking like a statistician, which is my serious advantage over Silver in the long run.

  1. Don't chase a single data point.  I was impressed by Buddy Roemer's results in one poll and changed my prediction because of it.  I should not do that and won't do it again.
  2. There's no point in being precise when you don't know what you are talking about.  The data I collect is to the nearest percent.  I shouldn't work with numbers to the nearest tenth of a percent.  My predictions will be to the nearest percent.  It giveth and it taketh away, but for the most part, it will giveth to me.  I might decided to round to the nearest half a percent, but no closer.  The data doesn't deserve it.
  3. Keep track of the recent, ignore ancient history.  Silver tweeted that every winner broke the 40% barrier since 1972.  This is because he's a statistician.  Completely meaningless number and somebody who isn't fooled by useless hard work could see through it in a fraction of a second.
There is no primary Tuesday.  The next test is a week from Saturday in South Carolina.  I will wait until I've checked the most recent polls on the day of the election but will post by blog call and my tweet long before the polls close. This is what Silver does, so it puts us on an even footing.

With no false modesty, I'm better than Nate Silver and I really do believe I will beat him at this in the long run.  I could use an old cliche and say I've forgotten more mathematics than he will ever know,  but once again with no false modesty, I haven't forgotten it.

He's a statistician. I'm a mathematician. I outrank him.


BobManDo said...

Awesome, Mathy Boy... er Matty Boy... So what contract with which of the Main Stream Media will you accept? Hold out for the big bucks... but please... don't go work for the Faux.

AphotoAday said...

LOVE this fragment from your text:
"The data doesn't deserve it."

Karlacita! said...

Wait, from a grammar azz: Isn't it "The data don't deserve it."?

You are to Silver as Mothra is to Osaka.

sfmike said...

Who's Nate Silver and why should we want you to stomp him, other than out of sheer friendly partisanship?

Matty Boy said...

BobManDo: Current TV has offices in San Francisco, so my plan is to go to them first, offer my services at what I consider a pittance and negotiate for later dates.

AphotoAday: I'm glad you liked it. I have a piece of your artwork and would be happy if you accepted one of mine, a geometric decoration.

Karlacita!: Your grammar correction is completely justified, but since I also went with "a single data point" instead of "a single datum", I'm sticking with the incorrect common usage.

SFMike: Nate Silver did stats with sports, then switched to politics. His blog became a big deal and the New York Times hired him. My predictions in 2008 were as good as his with a lot less work and a lot better grasp of math. I want to continue the ass-kicking this year, but I must admit doing predictions in a large field like this is new to me.

I need to get all Zen and shit and trust my superior knowledge of math instead of constantly trying to fiddle like Silver does. As people drop out, I fully expect my work to be better. When it gets to the general election, which will either be a two or three person race, my methods are significantly better.

Anonymous said...