This blog is still alive, just in semi-hibernation.
When I want to write something longer than a tweet about something other than math or sci-fi, here is where I'll write it.

Wednesday, January 4, 2012

Results for the first prognostication:
Not so good.


I picked the order Paul, Romney, Santorum at the top of the Iowa Caucuses.  It actually came out Romney and Santorum in a near dead heat for first and Paul many percentage points behind. I got the order of the also rans right, but missed their percentages.  Let's go through the wreckage candidate by candidate.

1st place: Mitt Romney 24.6% 
My prediction of 22% was off by 2.6%.

2nd place: Rick Santorum 24.5% 
I predicted 19%, my big screw-up of the evening.  I lose 5.5%.

3rd place: Ron Paul 21.4% 
I was 1.6% too high with 23%. 

4th place: Newt Gingrich 13.3% 
Too high at 15%. I lose 1.7%

5th place: Rick Perry 10.3% 
Much closer.  Off by 0.3%

6th place: Michele Bachmann 5% 
Once again, two points too high at 7%. 

7th place: Jon Huntsman 0.6% 
I thought he'd get about 3%.  I overestimated the sanity of the Iowa caucus goer. I lose 2.4%

Others: 0.3% 
I said 1%.  Take the 3% I over-estimated Huntsman and the low end choices and give it to Romney.

What grade do I deserve?  If we accumulate the percentage mistakes I made, I get an 83.9%, which I would call a B.  I think that's a grade I can live with.
As for the places, There are eight positions, so if we say the basic adjustment is to switch two adjacent places, like second and third trading places, It would take sixteen such transpositions to go from completely backwards to correct.  To make my list right, I need to switch Paul with Romney for the top spot, then switch Paul from second to third with Santorum. 14 right out of 16 would be 87.5%, about on the cusp of B and B+.

Of the two grading systems, I think the accumulated percentage mistakes is the fairest.  As people drop out, getting the correct order of finish will get easier, though when it gets down to two candidates, then it becomes pass/fail.

So, unlike many other prognosticators, I will keep track of my grades.  Could we lose a couple candidates soon?  Signs point to yes. Perry slunk back off to Texas last evening, for example, and reality might even seep through Michele Bachmann's caked on make-up, touching the virgin territory of her brain.  Every person who leaves the race makes my job slightly easier.

And just to make myself feel a little better, I took the last seven polls taken in Iowa, adjusted the "None of the Above" down to zero, increased all the other percentages proportionally and took their grades on the caucuses. Here are the names of the companies, the date of the polls and their grades.

Rasmussen(12/28): 82%
NBC News/Marist (12/27-28): 81%
Des Moines Register (12/27-30): 81%
PPP (12/31-1/1): 80%
We Ask America (12/29): 74%
ARG (12/29-1/1): 68%
Insider Advantage (1/1): 62%

As you can see, polling as close to the actual date wasn't much help.  Insider Advantage flat out sucked asking people on New Year's Day, while Rasmussen did the best when they packed up their tent on December 28.  Everybody underestimated Santorum.

My next prediction will be next Monday before New Hampshire votes on Tuesday.

2 comments:

namastenancy said...

I read that the count may be off so you may be right after all.

Matty Boy said...

No, it can't fix all my troubles. My only solace is that when counting to the nearest tenth of a percent, I beat Nate Silver's prognostication.