**7 Dec GA Senate runoff model update -** Thinking about [this tweet](https://twitter.com/Nate_Cohn/status/1335969234413883395), I will be making the 3 Nov election a much larger portion of my GA runoff model (adding more weight to the prior). This pushes the odds I give for Ds taking the Senate from 40% to 35%. 538 [now has a polling average for GA](https://projects.fivethirtyeight.com/georgia-senate-polls/). Even though they have the same polls, they say Ossoff +0.3 and Warnock +2.1 whereas I say... w/ election prior - Purdue +1.9, Loeffler +0.5 wi...
I polled California Proposition 16 on 20 October and found 46.1% topline support (N=869 likely California voters, weighted to be representative of California by race, age, gender, education, income, and 2016 Presidential vote). Using information from this poll, plus historical data on the accuracy of polls, but no other information (e.g., other polls, opposition spending), I model an 80% chance that the true support as observed on election day will fall 30.4% and 49.4%. A Monte Carlo simulation of the above 80% interval as a normal distribution with 10...

Per this super simple model (now with t-distributions!), as of 16 October, Trump has a 4.8% chance of winning https://www.getguesstimate.com/models/16846

— edited by PeterHurford

Here's the number of faithless electors in past elections:

2016 7

2012 0

2008 0

2004 1

2000 0

1996 0

1992 0

1988 1

Note that since 2016 there has been increased legislation to bind electors, but I still wouldn't be too surprised if some electors defected from Trump.

— edited by PeterHurford

So where is everyone from? I'm in Chicago

@Linch In fact, there's a decent amount of circumstantial evidence that the Metaculus prediction is usually worse than the community prediction.

@Sylvain a seventh example of surprisingly high Metaculus formula odds (relative to the community median and the presumed resolution).

UPDATE: At this point, Michigan, Nevada, North Carolina, Alaska, and Pennsylvania are well out of the range of being contested that I give them each a <0.1% chance of flipping (would require some very unprecedented event like an extreme norms violation or discovery of widespread fraud). Wisconsin, Arizona, and Georgia could conceivably be turned over in a recount but it would be unprecedented for a recount to overturn a margin of that size. Maybe these are 0.5% each. Overall I expect these to be uncorrelated, so the end probability is `1-((1-0.001)^5*(...

@Natalia_Mendonca @casens my current predictions have been relying on a revelation of a secret pardon after January 20 not counting for positive resolution. I think others have also predicted this way. I think a permanently negative or ambiguous resolution makes the most sense here. Maybe ambiguous since others may be thinking that a secret pardon reveal does count.

I'd love to have another Metaculus question about whether a secret pardon is revealed.

4 Dec GA Runoff Model Update

I took the new Rasmussen poll and changed my model to a weighted analysis of polls with the election as a prior and now I think the Dems have a 41% chance of taking the Senate :O

Rasmussen polls are usually biased towards Republicans so them finding Ossoff and Warnock in the lead is a big deal.

Also the SurveyUSA poll from earlier is a big boost and hard to shake off.


WHO calls for emergency meeting on new virus in China, as cases spread to health care workers https://www.statnews.com/2020/01/20/who-calls-for-emergency-meeting-on-new-virus-in-china-as-cases-spread-to-health-care-workers/ "Flash-forward to Tuesday: the number of cases of 2019-nCoV, as the virus is known, has leaped to nearly 300 and there are sick people in at least five other countries. That includes the first case in the United States." https://www.vox.com/2020/1/21/21075017/sars-wuhan-pneumonia-coronavirus Trump on coronavirus case found in US: “w...

There are also many other rumored pardons we are still waiting on, so I don't think we've seen the last of this.

Bevin, the ultimately unsuccessful candidate for the 2019 Kentucky governor election, didn't concede for nine days.

Does anyone want to make a question for September 12?

@Tamay Seems even harder to beat the Metaculus prediction when it can predict <1%, though perhaps I shouldn't be trusted with the ability to lose tens of thousands of points.