The aspect of this question that gives me the most trouble is that the "Thinking Fast" part of the brain tells me that Trump is going to win, but the "Thinking Slow" part says, nah, not a chance. Example: I read that Trump supporters won't bother going to the caucuses because it's too hard. My "Thinking Fast" function says, those pundits are whistling in the dark. My"Thinking Slow" part says, Trump supporters don't have a history of getting off their rear ends, as supported by both statistics (apparently) and anecdotes. Then my "Thinking Fast" function shouts at me, that's biased by the contempt the intelligencia has for white, lower income males, and pundits who habitually know how to lie with statistics. Thinking Slow part argues, how do I know the intelligencia is so prejudiced against Trump supporters? How do I know what the true demographics of them might be?
To date, I have given up trying to wade through my own biases and interpretations of others' biases around the election, and have knee-jerk relied on betting markets. Those have their own problems, but I think lend themselves much more to a quantification of Trump's (and others') chances than trying to come up with an appropriate weighting of those who are certain Trump will win (based on far-out polls, press-coverage, perceived momentum, etc.) and those who are certain (they hope!) that he can't. Betting markets (and aggregators of markets) appear to be more level-headed in interpreting evolving conditions thus far.
As elections get closer and top poll-analysts more confident, I will possibly switch from betting markets to good analysts, or a blend of the two.
They go into much detail about the difficulties of massaging prediction poll data sufficiently so as to make it as good or better than prediction market data. So why, then, in their attempt to commercialize their research, are they running GJopen in pure prediction poll mode?
@cmeinel, good reference. They claim that with fewer forecasters, and when forecasters for whatever reason don't communicate, and when it is necessary to compartmentalize information, polls are more practical. So we can forget about the market issue. For markets I guess predictit.org is the current research platform. I've played on that platform but I didn't find it very engaging, and the rake is so high that it is not tempting to stay and adapt.
Although having updated my forecasts for the market research we are doing on most popular toys, I really miss the ability in market format to trade out of a question which is utterly unappealing. I now regret entering a bunch of questions that I have no interest in.
This actually is a pretty darn good analysis of Trump's strategy. Raising probability.
Dateline 12-27-15 http://blog.dilbert.com/post/136042658956/is-iowa-a-caucus-or-a-mental-health-problem
OK I'm reblogging that. [1]
[1] https://larswericson.wordpress.com/2015/12/28/models-read-my-book/
Dilbert is sometimes funny, but Scott Adams is the worst.
Here are decent looking betting lines (albeit with an obscene ~25% vig, so better to use as approximate % chances rather than precise ones): http://www.paddypower.com/bet/politics/other-politics/us-politics?ev_oc_grp_ids=2092272
The aspect of this question that gives me the most trouble is that the "Thinking Fast" part of the brain tells me that Trump is going to win, but the "Thinking Slow" part says, nah, not a chance. Example: I read that Trump supporters won't bother going to the caucuses because it's too hard. My "Thinking Fast" function says, those pundits are whistling in the dark. My"Thinking Slow" part says, Trump supporters don't have a history of getting off their rear ends, as supported by both statistics (apparently) and anecdotes. Then my "Thinking Fast" function shouts at me, that's biased by the contempt the intelligencia has for white, lower income males, and pundits who habitually know how to lie with statistics. Thinking Slow part argues, how do I know the intelligencia is so prejudiced against Trump supporters? How do I know what the true demographics of them might be?
I play poker with a Trump supporter who is a retired economics professor from South Carolina who drives a new Lincoln MKZ, so that's a data point.
@cmeinel
Well said.
To date, I have given up trying to wade through my own biases and interpretations of others' biases around the election, and have knee-jerk relied on betting markets. Those have their own problems, but I think lend themselves much more to a quantification of Trump's (and others') chances than trying to come up with an appropriate weighting of those who are certain Trump will win (based on far-out polls, press-coverage, perceived momentum, etc.) and those who are certain (they hope!) that he can't. Betting markets (and aggregators of markets) appear to be more level-headed in interpreting evolving conditions thus far.
As elections get closer and top poll-analysts more confident, I will possibly switch from betting markets to good analysts, or a blend of the two.
I think in the GJ vault they know market format works better. Forecasting format is a compromise to avoid the FutureMap problem.
@000 I've been puzzling over the market vs prediction polls accuracy issue, too. See the GJP team's latest paper at https://www.researchgate.net/publication/281765164_Distilling_the_Wisdom_of_Crowds_Prediction_Markets_versus_Prediction_Polls
They go into much detail about the difficulties of massaging prediction poll data sufficiently so as to make it as good or better than prediction market data. So why, then, in their attempt to commercialize their research, are they running GJopen in pure prediction poll mode?
@cmeinel, good reference. They claim that with fewer forecasters, and when forecasters for whatever reason don't communicate, and when it is necessary to compartmentalize information, polls are more practical. So we can forget about the market issue. For markets I guess predictit.org is the current research platform. I've played on that platform but I didn't find it very engaging, and the rake is so high that it is not tempting to stay and adapt.
Although having updated my forecasts for the market research we are doing on most popular toys, I really miss the ability in market format to trade out of a question which is utterly unappealing. I now regret entering a bunch of questions that I have no interest in.
@cmeinel...CHAMPS KNOW. CHAMPS KNOW CHAMPS KNOW CHAMPS KNOW. OK now I should remember that excellent mnemonic for at least another week or two.