HomeLatest ThreadsGreatest ThreadsForums & GroupsMy SubscriptionsMy Posts
DU Home » Latest Threads » Loki Liesmith » Journal
Page: « Prev 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Next »

Loki Liesmith

Profile Information

Member since: Thu May 26, 2016, 09:07 AM
Number of posts: 4,427

About Me

God of lies. Like math.

Journal Archives

Election Model Update 9/25

9/24 Update

9/21 Update

9/22 Update

9/1/2016 update

8/16/2016 update

Update with drift included 8/6/2016

Original Model Writeup 8/3/2016

Model projections:

#############################################
# INSTANTANEOUS MODEL
#############################################

Instantaneous Probability of a Clinton Win: 71.25 %
Mean Electoral Votes Clinton Currently Winning: 291.57
Median Electoral Votes Clinton Currently Winning: 294.0
Maximum Likelihood Electoral Vote Scenario for Clinto: 299

#############################################
# PROJECTING RESULTS TO NOVEMBER
#############################################

Probability of a Clinton Win if Current Trends Continue: 63.29 %
Mean Electoral Votes Clinton Projected to Win: 284.25
Median Electoral Votes Clinton Projected to Win: 285.5
Maximum Likelihood Projected Electoral Vote Scenario: 269

Predicted National Point Spread (Clinton = +, Trump = -) 2.45
Posted by Loki Liesmith | Sun Sep 25, 2016, 01:36 PM (6 replies)

538's estimates of probability net us almost ZERO INFORMATION

Just look at this graphic of the national poll average and the 538 probabilities



Maybe, MAYBE the 538 model is squeezing some information out at the margins. I bet I could construct a model that does as well or better than 538 simply by taking the differences in the two poll averages and converting it to a direct probability.

Posted by Loki Liesmith | Sun Sep 25, 2016, 01:07 PM (49 replies)

Any movement in this race over the last three weeks is an illusion

Look at this graph of poll outcomes vs when they are in the field:

:large

legend:
Position of poll on y axis is size of Clinton Lead.
Intensity of color of boxes is number of polls in field on that interval.
Length of boxes is the period over which the poll is taken.


https://twitter.com/williamjordann/status/780032565176508416
Posted by Loki Liesmith | Sun Sep 25, 2016, 11:05 AM (3 replies)

Sample error

Last week same poll had her up 8.
Conclusion: she's up 5.
Posted by Loki Liesmith | Sun Sep 25, 2016, 07:49 AM (0 replies)

Michigan would be my only worry at this point

Been a while since we've had fresh data from there.

National polls are instructive in that they are highly correlated with movements in many of the states.
Posted by Loki Liesmith | Sun Sep 25, 2016, 02:51 AM (2 replies)

Washington Post Poll

https://www.washingtonpost.com/politics/poll-clinton-trump-in-virtual-dead-heat-on-eve-of-first-debate/2016/09/24/b99c95de-81cb-11e6-8327-f141a7beb626_story.html

H2H: C49 T47
4way C46 T44 rest: who cares

Bad points: obviously closer than last post poll.

Good points: Clinton is near 50 in the two way. Registered voters are split evenly. That's almost certainly NOT the case. Also respondents don't expect Clinton to win the debate by that large a margin 44% expect her to in, 34$% expect Trump to win. Expectations game may be a wash.

Polls from Thurs/Friday put this at around a 6 point race. Average with this, looks like a 4-5 point race at this point. Even the write up above acknowledges other polls disagree with this one. They know they have a slightly funky sample.
Posted by Loki Liesmith | Sun Sep 25, 2016, 02:30 AM (25 replies)

Wanna build your own likely voter model

Upshot is going to help you:

https://twitter.com/Nate_Cohn/status/779342214447104000
Posted by Loki Liesmith | Sat Sep 24, 2016, 10:08 AM (1 replies)

Election Model Update 9/24//2016

9/21 Update

9/22 Update

9/1/2016 update

8/16/2016 update

Update with drift included 8/6/2016

Original Model Writeup 8/3/2016

Model projections:

#############################################
# INSTANTANEOUS MODEL
#############################################

Instantaneous Probability of a Clinton Win: 72.3629756738%
Mean Electoral Votes Clinton Currently Winning: 291.8536
Median Electoral Votes Clinton Currently Winning: 293.0
Maximum Likelihood Electoral Vote Scenario for Clinto: 297

#############################################
# PROJECTING RESULTS TO NOVEMBER
#############################################

Probability of a Clinton Win if Current Trends Continue: 63.2426988922%
Mean Electoral Votes Clinton Projected to Win: 283.249
Median Electoral Votes Clinton Projected to Win: 284.0
Maximum Likelihood Projected Electoral Vote Scenario: 281

Predicted National Point Spread (Clinton = +, Trump = -) 2.45106856654

Analysis: Appears that recent solidification in polls has boosted Clinton's numbers in the intantaneous projection. The extrapolated odds at election day still look about the same.

Clinton Still Favored.
Posted by Loki Liesmith | Sat Sep 24, 2016, 09:08 AM (5 replies)

How the Trumpists are trying to game debate expecations

Lowering Expectations:

There is a conventional wisdom hardening which holds that Hillary Clinton’s most urgent task at the first debate is to bait Donald Trump into acting like the ignorant, dangerous, hateful, bigoted, temperamentally unhinged character he has periodically displayed to the nation for the last year. The corollary to this is that if Trump acts relatively controlled and projects minimal seriousness, he might “defy expectations” and emerge the winner.

I don’t buy it. While this is certainly a factor worth considering, I’m going to suggest that Clinton can win this debate — in the minds of voters, if not in the minds of pundits — even if Trump does pull off that magic transformation for ninety minutes on Monday night.

No question, Team Clinton probably would like to see Trump come unglued before an audience that could be as large as 100 million people. The New York Times reports that Team Clinton is trying to determine “how to knock Mr. Trump off balance,” in the belief that “she needs the huge television audience to see him as temperamentally unfit for the presidency, and that she has the power to unhinge him.”

And no question, Team Trump believes that avoiding that outcome is key to his hopes of prevailing. The Associated Press reports that Trump’s advisers are counseling him to avoid letting Clinton rile him up, noting, remarkably, that “some Trump aides are more concerned about Trump’s disposition on the debate stage than his command of the issues.”


A tweet/tag to retweet if you want to help remind the media to do it's job:


https://twitter.com/lambertglowbug2/status/779335252204457985
#LoweringExpectations
Posted by Loki Liesmith | Fri Sep 23, 2016, 11:24 AM (23 replies)

Election Model Update 9/22/2016

#############################################
# INSTANTANEOUS MODEL
#############################################

Instantaneous Probability of a Clinton Win: 0.684513006655
Mean Electoral Votes Clinton Currently Winning: 285.9956
Median Electoral Votes Clinton Currently Winning: 288.0
Maximum Likelihood Electoral Vote Scenario for Clinto: 287

#############################################
# PROJECTING RESULTS TO NOVEMBER
#############################################

Probability of a Clinton Win if Current Trends Continue: 0.627906976744
Mean Electoral Votes Clinton Projected to Win: 283.764
Median Electoral Votes Clinton Projected to Win: 284.0
Maximum Likelihood Projected Electoral Vote Scenario: 270
Predicted National Point Spread (Clinton = +, Trump = -) 3.03117822284

Posted by Loki Liesmith | Thu Sep 22, 2016, 10:47 AM (0 replies)
Go to Page: « Prev 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Next »