TIA analysis: June State and National poll results CONFIRM each other.

[TruthIsAll]

This is interesting. I have calculated the Kerry/Bush popular
vote percentages, based on the latest state polls (mostly
taken in June), weighted by the electoral votes for each
state. And I assumed a 50/50 undecided split. 

The resulting popular vote split is EXACTLY 50.0/50.0. This
confirms the 13-poll June averages (50.6 Kerry/49.4 Bush)
which are adjusted by the same (CONSERVATIVE) 50/50 undecided
split. 

But here's the good news:
1) Kerry will get near 70% of the undecided, based on
historical voting analysis.

2) Even with a VERY conservative 50/50 split in the POPULAR
vote, the electoral vote probability simulation shows that
Kerry has an 80% chance of winning, because the electoral vote
distributions work in his favor (he wins the big states, Bush
has big leads in SMALL states). 

3) With a current 52-48 split in the popular vote,
Kerry/Edwards is beginning to open up daylight and may be
pulling away from Bush. A November landslide is not out of the
question.

Bottom line, the independent analysis of state, national and
Electoral simulation model projections confirm each other.
  
13 -POLL AVERAGE		Undecided Adj (50/50)	
Month	Kerry	Bush 	Diff	Kerry	Bush 	Diff
Jan 	40.8	51.6	-10.8	44.6	55.4	-10.8
Feb	47.4	46.0	1.4	50.7	49.3	1.4
Mar	47.4	44.7	2.6	51.3	48.7	2.6
Apr	46.5	45.3	1.2	50.6	49.4	1.2
May	46.5	44.4	2.1	51.0	49.0	2.1
June	46.9	45.6	1.3	50.6	49.4	1.3 <<<<<
July	49.0	45.0	4.0	52.0	48.0	4.0
						


Here is the simulation run based on the state polls.

Undecided to Kerry: 50%	
SIM	EV			
1	279	Win		
2	294	Win		
3	303	Win		
4	311	Win		
5	325	Win		
6	269	Tie		
7	278	Win		
8	265	Lose		
9	294	Win		
10	299	Win		
AVG	292			
Win	8			


State	EV	AdjK	Diff	Prob(win)
AL	9	40.5	(19.0)	 0.001 
AK	3	38.5	(23.0)	 0.001 
AZ*	10	44.0	(12.0)	 0.001 
AR*	6	51.0	2.0 	 0.635 
CA	55	57.0	14.0 	 0.990 
CO	9	47.5	(5.0)	 0.202 
CT	7	59.0	18.0 	 0.990 
DE	3	56.5	13.0 	 0.990 
DC	3	90.0	80.0 	 0.990 
FL*	27	52.5	5.0 	 0.807 
GA	15	41.5	(17.0)	 0.001 
HI	4	59.5	19.0 	 0.990 
ID	4	35.0	(30.0)	 0.001 
IL	21	56.5	13.0 	 0.990 
IN	11	42.0	(16.0)	 0.001 
IA*	7	52.0	4.0 	 0.750 
KS	6	40.0	(20.0)	 0.001 
KY	8	43.5	(13.0)	 0.001 
LA	9	47.0	(6.0)	 0.145 
ME	4	50.5	1.0 	 0.577 
MD	10	57.0	14.0 	 0.990 
MA	12	64.5	29.0 	 0.990 
MI*	17	52.5	5.0 	 0.807 
MN*	10	52.5	5.0 	 0.807 
MS	6	34.5	(31.0)	 0.001 
MO*	11	49.0	(2.0)	 0.375 
MT	3	40.0	(20.0)	 0.001 
NE	5	35.5	(29.0)	 0.001 
NV*	5	51.5	3.0 	 0.692 
NH*	4	51.5	3.0 	 0.692 
NJ	15	55.0	10.0 	 0.990 
NM*	5	53.5	7.0 	 0.923 
NY	31	59.5	19.0 	 0.990 
NC	15	46.5	(7.0)	 0.087 
ND	3	36.0	(28.0)	 0.001 
OH*	20	48.0	(4.0)	 0.260 
OK	7	40.5	(19.0)	 0.001 
OR*	7	53.5	7.0 	 0.923 
PA*	21	52.5	5.0 	 0.807 
RI	4	61.0	22.0 	 0.990 
SC	8	42.5	(15.0)	 0.001 
SD	3	42.0	(16.0)	 0.001 
TN*	11	41.0	(18.0)	 0.001 
TX	34	41.0	(18.0)	 0.001 
UT	5	27.5	(45.0)	 0.001 
VT	3	57.5	15.0 	 0.990 
VA	13	47.5	(5.0)	 0.202 
WA*	11	54.0	8.0 	 0.975 
WV*	5	47.0	(6.0)	 0.145 
WI*	10	48.0	(4.0)	 0.260 
WY	3	30.0	(40.0)	 0.001 
	538			
				

[TruthIsAll]

1. I will be posting this analysis on a regular basis.

Using the following methodology:

Assume a two-man race.
1) The 13 national poll numbers are adjusted by allocating the undecided vote to Kerry. I assume it to be 70% based on historical statistical studies.
2) Calculate monthly trend averages of national polls.
3) Run Monte Carlo random number simulations of state polls (+/-4%) Margin of Error to calculate the probability of Kerry winning.

[trogdor]

3. For non-CS majors.

A Monte Carlo algorithm is one in which an otherwise (mathematically) unsolvable problem has lots and lots of numbers thrown at it by a computer running a random number generator. Most of the results will cluster around the solution set, and the accuracy of the result increases with the number of tries.

[TruthIsAll]

4. You are talking about Genetic Algorithms ...

GA is used to determine the "best", optimal result by running lots of trials and eliminating "bad" offspring results. This is "survival of the fittest" in Darwinian Theory.

That is NOT what I am doing here.

I am using Monte Carlo simulation, a method which uses probabilities of events and applies random numbers to generate many outcomes from which we can surmise a statistical probability.

In this case, the "event" is winning the state's popular vote and the outcome is the combination of states which Kerry wins. Once we have the states, we can then calculate the total electoral votes. If we run the simulation 10 times and Kerry's total exceeds 269 electoral votes 9 of the 10 times, we can estimate that he has about a 90% probability of winning the election.

Kerry has a defined probability of winning each state, for example:

Kansas: .001
New York: .999
Florida: .807

The probabilities are determined as a function of the polling numbers. For instance, say the latest Kansas poll is 60% for Bush, 40% Kerry. Kerry probably has a 1 in 1,000 chance of winning the state.

On the other hand, if Kerry is ahead by 5% in Florida, then he has a 81% chance of winning the state.

A random number generator returns a value X between 0 and 1. If X is less than the probability of Kerry winning the state, we give the state to Kerry, otherwise it goes to Bush.

For instance, if X = .72 for Florida, then Kerry wins the state. The random number X= .72 is LESS than the .807 probability that Kerry will win FL given that the polls have him leading by 5%.

We do this simultaneously for all the states, and sum the electoral votes for those which fall in Kerry's column.

[Ishoutandscream]

2. TIA, several months ago I thought you were a bit premature on your

predictions. But I'm right there with you now. We're going to win this thing!!