This chart tells it ALL: Kerry has rocketed back.

[TruthIsAll]

The Kerry rebound over the last two days as reflected in the latest Zogby and ARG state polls is EXTREMELY significant.




Of course behind the climb in Electoral vote is the rise in popular vote.



Blue is Good.



And for you, Dan Rather, what's the frequency, Kenneth?


Want more of this?
http://www.geocities.com/electionmodel/

[Carla in Ca]

1. and lets keep it that way! eom

[Catch22Dem]

2. Hey TIA

What, exactly, is the Monte Carlo simulation? I took stats in college, but don't remember much about anything! HAHAHA

[lefty_WOHM]

9. Here's a site w/ a quick description

http://www.decisioneering.com/monte-carlo-simulation.html

Here's another:

http://www.chem.unl.edu/zeng/joy/mclab/mcintro.html

But TIA will have to explain how exactly he uses MC.

[Catch22Dem]

11. thanks! (n/t)

[TruthIsAll]

14. Thanks. I get tired talking about it. I should have seen this.

tia

[TruthIsAll]

10. Monte Carlo simulation is the methodology used

to simulate a system using random numbers applied to the probabilities of system events. The simulation is run many times to determine the frequency of possible outcomes.

In other words, if there are many uncertain events (like who will win each states electoral votes) we use the latest poll numbers to assign a probability of winning each state to Kerry.

A random number between ZERO and ONE is generated and compared to the probability of winning the state. If the RND is less than the probability, Kerry wins the state and gets the electoral votes.

For example, Kerry is leading in MN 47-45. Assuming he gets 60% of the undecided vote, his probability of winning the state is 81% (this is my base case). If he gets 50% of the vote, the prob is 70%.

If the random number generated is less than .81, Kerry wins the state; otherwise, Bush wins.

We do this for all 50 states and DC. and total up the EV's. That is ONE trial run. I run 5000 of them and determine how many of them Kerry won (4694). So he has a 93.9% probability of winning (4694/5000). I also calculate his average EV (306) for the 5000 trials.
I run the model for the five cases: 50,55,60,67,75% of the undecided to Kerry. That's a total of 5 * 5,000 = 25,000 simulation trials.

Hope that clears it up.

tia

[Catch22Dem]

12. Thanks! (n/t)

[mconvente]

26. nice

what programming language do you use to run these tests?

[TruthIsAll]

32. This is just a plain old Excel workbook.

I use the RND function to generate random numbers and the NORMDIST function to compute probabilities.

[coloradodem2005]

3. I printed that first chart.

;)

[mdguss]

6. It's good:

ARG surveyed 600 people by phone in ALL 50 states. That's a very, very expensive poll that no one else will do. It's more valid than a Gallup Poll that surveys 500 people nationwide. ARG shows Kerry on the cusp of victory. But we must GET OUT THERE and campaign for Kerry to win.

[UCLA Dem]

4. Beautiful. You just made my week!!!

[meti57b]

5. Thank you!

[renoray]

7. How is that figured?

It would nice to see Kerry with a 120% chance of winning and all, but those numbers don't make much sense to me. If it is by electoral votes, what states were doing those switches over the past week?

[TruthIsAll]

17. Look at the poll report which follows the graphs here..

http://www.geocities.com/electionmodel/

[lefty_WOHM]

8. OMG! You made my day!

You rock! :headbang:

[goclark]

18. Let's Rock Around The Clock!

:toast:

[foo_bar]

13. hubris, Achilles

[TruthIsAll]

15. Right.

tia

[Nicholas_J]

16. Both Zogby and ARG

Have Kerry leading in Electoral Votes. Zogby has Kerry with a predicted 296, but ARG has Kerry with an ACTUAL 270 Electoral votes right now, just the amount needed to win the election.

[TruthIsAll]

19. I show projections based on the current Zogby and ARG polls

and assuming Kerry gets 60% of the undecided. That's not a stretch at all.

In fact, it is quite conservative, because Kerry will do better than that (75%) and also because the polls DO NOT reflect his true underlying strength with those who are newly registered and are not considered likely voters...

[Nicholas_J]

20. Those are the figures that apply to the last 3 weeks

of a presidental campaign. DUring the last two to three weeks of a presidential campaign 75 to 85 percent of those left tend to go againt the incumbent. The incumbent is a known factor, adn anyone who has not made up their mind by that point has not done so becasue they have misgivings about the incumbent and are waiting to find out as much as possible about the opposition.

[RobertDevereaux]

21. Splendid! n/t

[DeepModem Mom]

22. Thanks, TIA!

[Andromeda]

23. What would we do without you TIA...

We'd be wringing our hands and chewing our nails, that's what!
:toast:

[VOX]

24. Indeed, Truth**Is**All!!!

Many thanks! :toast:

[TruthIsAll]

25. I did NOT update the model today. Here's why..

Electoral-vote.com today included new polls from Repub leaning state pollsters which wipe out some of the gains from Zogby and ARG.

I do not believe these polls. Why mess up the garden of growing support for Kerry with them? These new polls are designed to obscure, not inform.

The latest Zogby and ARG poll numbers are from professional, non-biased organizations I trust. So they are staying in for a while longer.

[demokatgurrl]

27. Please explain to dumb lawyer...

Why these numbers look so good, e.g. probability of winning, when www.electoral-vote has looked so bad lately?

[TruthIsAll]

29. Electoral-vote.com has Kerry in front, but not as much as my model.

I use Electoral-vote's numbers, but my methodology is different.

I assume Kerry will win 60% of the undecided vote, for one thing. And I calculate probabilities of a Kerry win using a 5000 election-trial Monte Carlo simulation.

Electoral-vote, like many election sites, gives a state's EV to whoever is leading. The Monte Carlo simulation method which I use awards the electoral votes to both, depending on the result of each election trial, of which there are 5000. For example, if Kerry has an 80% probability of winning a state, he'll win 80% of the trials, Bush will win 20%.

Electoral-vote.com today replaced good polls just out from ARG and ZOGBY with Repub pollsters that I do not trust. For this reason, I will not update the data today.

[zaj]

28. Is it me or does this chart show...

... that the entire idea of using m.c. simulation to predict the probability of a Kerry win is just worthless?

It feels good to have someone tell me that Kerry has a 97% chance of winning, but when polling number moves over the course of a week or two can swing the Kerry-Win probability from 97% down to 18% and back up to 97% in the span of 10 days...


That suggests to me that the tool is pretty worthless for anything other than a feel-good effect.

[TruthIsAll]

30. If you look at the trend graphs, you will see that the probs were pretty

constant for Kerry in the 90% range in July and early August.

His drop in the polls was steady over a one month period. But the latest polls reversed the trend and put him in the lead, so his pop. vote percectage, electoral votes and probability all rose accordingly.

Going up 2% in the national polls, as Kerry did last week, is a major move. In any case, at 94% he can't climb more than another 6%.

He will get to 99.99% when he has reached 335 electoral votes.

You must look at the trends, understanding that every once in a while there is a rapid trend reversal.

ON EDIT: IT IS NOT MY INTENT TO USE THE MONTE CARLO/GRAPHIC ANALYSIS TO "FEEL GOOD". IT IS THE BEST WAY TO DESCRIBE THE ODDS OF A KERRY WIN, EVEN AS THEY CHANGE FROM DAY TO DAY. IT'S SURELY PREFERABLE TO JUST ASSIGNING ELECTORAL VOTES BASED ON A ONE POINT LEAD IN A QUESTIONABLE STATE POLL.

If Kerry has 52%-48% in the final polls before election day, it's good to know the probability just in case Bush tries to Diebold him.

[zaj]

33. I guess my point is that...

... the simulation is so sensitive to small poll movements that even movements less than the margin of error can create wild swings in the probability of a Kerry win.

It seems that it would be almost impossible to gain any meaningful information from such a tool.

I would like to clarify that I do think your site and your activities are very cool. I am all for new ways of looking at election results. I am interested in your work, and I am thankful that you share it here.

But I just wanted to discuss the limitations of the practice.

:toast:

[TruthIsAll]

34. Sorry to disagree. The simulation and national poll average

is NOT sensitive to changes in the polls; the averaging effect and the result of 5000 simulations solves that problem.

Over the last few days, there has been a major turnaround in multiple state polls from Zogby and ARG. This is a phenomenon which has never occurred in the 10 weeks that I've been running the model.

Look at the trends in pop. vote, electoral vote and win probability -they have been uniform and smooth except for the sudden Kerry move over the last few days in these two polls.

By contrast, I would point you to sites which, like electoral-vote.com, see EV's change sharply day-to-day. For example, a one point switch in a single poll (say Florida) would cause a 27 point change in EV. This would NOT happen in the Monte Carlo simulation I run, because the change would only impact the probability of winning the state.

For example, assume a 55% probability for a Kerry win in FL since he leads 51%-49%. The next day the lead changes to 51-49 Bush. Bush would win very close to 55% of the 5000 simulated Florida trial elections and Kerry would win 45%. As a result, the average Kerry total EV would not decline by 27 EVs. It would drop by just a few EV's.

The results are much less sensitive to day-to-day poll changes. The fact is that between Zogby and ARG, virtually all the battleground states moved in Kerry's favor in two successive days, which resulted in the Kerry spike.

[Night Gal]

31. Thanks - made me feel much better

I'll stick to reading YOUR polls from now on

[autorank]

35. Truth, you're the 'person' (PC). n/t/

[bush equals idiot]

36. That's not all.

I keep saying the surprises haven't even come yet. Overseas ballot requests are so huge, the Pentagon tried to cut them off by shutting down the registration website. Also the release of Farenheit coming up. And the historic numbers of superstar rock bands concerts shortly in all the battleground states.

[DemBones DemBones]

37. Deserves a few good kicks for the morning crew.

:kick:

:kick:

:kick: