tag:blogger.com,1999:blog-4052593945054595675.post3983013446105654953..comments2024-03-28T14:16:10.498+11:00Comments on Dr Kevin Bonham: Poll Roundup and Seat Betting Watch: Ghost of '75 Edition (20 August - Updates Added)Kevin Bonhamhttp://www.blogger.com/profile/06845545257440242894noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-4052593945054595675.post-815180947653891922013-08-21T22:37:27.987+10:002013-08-21T22:37:27.987+10:00As I understand it the Pottinger model isn't b...As I understand it the Pottinger model isn't based off individual seat markets; I believe it models betting market expectations off the headline. Kevin Bonhamhttps://www.blogger.com/profile/06845545257440242894noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-62612407825433719882013-08-21T20:36:56.642+10:002013-08-21T20:36:56.642+10:00The Pottinger model isn't overconfident becaus...The Pottinger model isn't overconfident because of seat-to-seat correlations. <br /><br />Kevin - I agree with your point about probabilities, but as I noted in the post, probabilities are not a linear thing. My model is predicting a very small chance of an ALP win vs the market prediction. But you don't need to inflate the width of the seat distribution by very much in order to recover the market probabilities. <br /><br />I could be more direct in combining my probabilities. The moment, the betting market is used indirectly in order to assess just how reliable it is based on past data (i.e. betting market --> 2PP --> seat distribution --> probability.) Another option is to go poll model --> 2PP --> seat distribution --> probability from poll model + probability from market --> overall probability. The question then becomes how you combine the probabilities... It's now clear how you do this in a robust way to me at the moment.Julian Kinghttps://www.blogger.com/profile/09999443797525672413noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-8502354531488543292013-08-21T11:30:04.096+10:002013-08-21T11:30:04.096+10:00I saw Pottingers post about the overconfidence of ...I saw Pottingers post about the overconfidence of the model (based upon individual seat markets) the other day.<br /><br />It strikes me that the seat markets are not independant of each other, and thus subject to systemic risk in the same way as the housing market is (i.e. loan risk models are not really built to deal with a nationwide drop in house prices of 20-30% as occurs in a sharp downturn like that experienced in some parts of the world during the GFC).<br /><br />The chance for Labor victory lies not in incremental changes seat by seat, but in a (increasingly unlikely) major event that shifts the national (or alternatively a significant state) vote by a large amount.<br /><br />intuitivereasonhttps://www.blogger.com/profile/15444634755480881972noreply@blogger.com