Support this website
I am using Huffpost's numbers for today (pre-debate) for tonight's election
prediction. These numbers much more strongly favor Obama than either
of the two other numbers I am using. Simon Jackman has an article that
explains
the model used to make these numbers.
Using a strong random number generator (described in yesterday's blog), I simulated the presidential election 100,000 times. The simulated results for today's Huffpost numbers are as follows:
Here are some raw numbers showing how much influence a given state has on the election results. What I have done is look at all of the cases in the model where a given state is won by a given candidate, and compare how winning or losing the states affects their chances of being in the White House next year this time:
The random number generator rolls a 1000-sided dice by grabbing two bytes from RadioGatun[32]'s bitstream. This gives us a number between 0 and 65535; should the number be 65000 or higher, the program discards the roll and tries again.
My simulation with Huffpost's numbers
Using a strong random number generator (described in yesterday's blog), I simulated the presidential election 100,000 times. The simulated results for today's Huffpost numbers are as follows:
Chance of winning election Obama 96.7% Romney 3.2% Tie 0.2%
By State
Here are some raw numbers showing how much influence a given state has on the election results. What I have done is look at all of the cases in the model where a given state is won by a given candidate, and compare how winning or losing the states affects their chances of being in the White House next year this time:
Obama gets CO Romney 1.72988% Tie 0.10658% Obama 98.1635% Romney gets CO Romney 5.41102% Tie 0.225566% Obama 94.3634% Obama gets FL Romney 0.00313372% Obama 99.9969% Romney gets FL Romney 4.64833% Tie 0.224706% Obama 95.127% Obama gets NH Romney 2.34891% Tie 0.153086% Obama 97.498% Romney gets NH Romney 4.54009% Tie 0.152856% Obama 95.3071% Obama gets NV Romney 2.49987% Tie 0.118921% Obama 97.3812% Romney gets NV Romney 5.67856% Tie 0.281542% Obama 94.0399% Obama gets OH Romney 0.894281% Tie 0.100098% Obama 99.0056% Romney gets OH Romney 25.8884% Tie 0.682143% Obama 73.4294%The most interesting thing for me is that the model has only one path to victory for Romney if Obama is able to get Florida. I have made a map of Romney's one and only 1-in-100,000 (according to Huffpost) path to victory for Romney without Florida, which can be seen here:
http://elections.huffingtonpost.com/2012/custom-presidential-election-map#nkadmnjannjkjenjnThen again, Huffpost's model says that Obama right now only has a 32% chance of getting Florida (Nate Silver and Nerd Wallet give Obama a slightly higher chance of getting Florida, but they both give Obama a lower chance of winning the election).
A note on the random number generator
The random number generator rolls a 1000-sided dice by grabbing two bytes from RadioGatun[32]'s bitstream. This gives us a number between 0 and 65535; should the number be 65000 or higher, the program discards the roll and tries again.
Once it gets a good roll, it takes the low three digits of the random number to get a number between 0 and 999. For example, if the number is 45,737, the program uses the number "737" for the dice roll. If Obama's chance-in-1000 of getting a state is lower than the roll, Obama gets the state, otherwise Romney gets it. Using the 737 example and Florida, since Obama has a 32% chance at Florida in Huffpost's model, Obama won't get Florida since he needs a 319 or lower to get it.
In order to reduce spam, comments for this entry are now closed