Not quite sure what the deal is with the Glug website ("an eccentric mix of wine, drinks, politics, food, sport and other interesting things", apparently), but I am grateful to it for making my particular obsession in life seem a little less unusual. In conjunction with Crikey, the site is running a competition that requires participants to nominate Labor’s percentage chance of victory in each of the state’s 89 electorates. Prizes include "superior Glug wines", Crikey subscriptions and "incredible bragging rights". Crikey’s Richard Farmer tells us that "early Crikey contest polling gives Labor the following percentage chance of victory in the 89 individual seats":
Albert 91%; Algester 93%, Ashgrove 91%, Aspley 91%, Barron River 65%, Beaudesert 16%, Brisbane Central 95%, Broadwater 65%, Bulimba 92%, Bundaberg 62%, Bundamba 93%, Burdekin 26%, Burleigh 74%, Burnett 19%, Cairns 82%, Callide 10%, Caloundra 39%, Capalaba 93%, Charters Towers 22%, Chatsworth 69%, Clayfield 54%, Cleveland 82%, Cook 88%, Cunningham 13%, Currumbin 48%, Darling Downs 10%, Everton 92%, Ferny Grove 94%, Fitzroy 91%, Gaven 39%, Gladstone 19%, Glass House 83%, Greenslopes 91%, Gregory 9%, Gympie 13%, Hervey Bay 75%, Hinchinbrook 23%, Inala 95%, Indooroopilly 62%, Ipswich 94%, Ipswich West 87%, Kallangur 91%, Kawana 54%, Keppel 73%, Kurwongbah 91%, Lockyer 14%, Logan 93%, Lytton 96%, Mackay 91%, Mansfield 88%, Maroochydore 19%, Maryborough 11%, Mirani 8%, Moggill 20%, Mount Coottha 91%, Mount Gravatt 91%, Mount Isa 89%, Mount Ommaney 87%, Mudgeeraba 52%, Mulgrave 81%, Mundingburra 85%, Murrumba 91%, Nanango 9%, Nicklin 9%, Noosa 38%, Nudgee 91%, Pumicestone 78%, Redcliffe 56%, Redlands 89%, Robina 22%, Rockhampton 93%, Sandgate 90%, South Brisbane 95%, Southern Downs 11%, Southport 80%, Springwood 88%, Stafford 93%, Stretton 91%, Surfers Paradise 13%, Tablelands 9%, Thuringowa 86%, Toowoomba North 66%, Toowoomba South 13%, Townsville 86%, Warrego 13%, Waterford 92%, Whitsunday 84%, Woodridge 95%, Yeerongpilly 92%.
The lightweights among you might prefer to try the old-fashioned method of tipping the total number of Labor seats through Centrebet, which is currently offering the shortest odds on 58. As my more dedicated readers will already know, my own personal tips are available for perusal and ridicule on the Poll Bludger election guide. My current estimation is that the Coalition will hold all the seats it won in 2004; the Liberals will gain Kawana and Mudgeeraba from Labor and Noosa from Labor-turned-independent MP Cate Molloy; the Nationals will gain Bundaberg from Labor and Gympie from independent "candidate" (last I heard) Elisa Roberts; the Liberals will lose their by-election gains of Chatsworth and Redcliffe to Labor, whereas the Nationals will hold Gaven. It turns out that my findings are in almost perfect accord with those of Glug (which differs from me only in that it has Labor’s odds in Mudgeeraba at over 50 per cent, and then just barely), and my projection of Labor seats is exactly equal to Sportsbet’s. Final tally: Labor 58, Nationals 18, Liberal 8, independent 4, One Nation 1.
In other news, it only came to my attention yesterday that Sky News is running a program called "Queensland Votes 2006" at 9.30pm EST each night this week. Last night’s instalment featured interviews with Bruce Flegg, Barnaby Joyce and Sean Parnell of The Australian. The highlight for mine was host David Speers putting it to Flegg that Liberal members (I leave it to you to imagine which ones) had told him the party’s polling had it on track to win at least 20 seats before the August 7 leadership change.
I missed the Queensland Votes 2006 show. Just out of curiousity, what was Flegg’s response to Speers’ question?
Maybe I’m stupid, but I can’t see how predicting the “possibility” of a seat, and matching that with the actual outcome would allow any meaningful ranking of the predictors. Why didn’t they just do “who will win every seat” if that’s what they wanted to do. Seems convoluted to me. Maybe too many people would have picked the same outcomes, I guess.
Interesting stats, I presume the seats to watch will be the 45-55% range. I say this is just because you think someone will win a seat doesn’t mean you will vote for them.
“I can’t see how predicting the “possibility†of a seat, and matching that with the actual outcome would allow any meaningful ranking of the predictors” – actually it is quite a valid predictor of the outcome, provided that you get a lot of people to give their opinions. It is analogous to the Delphi Technique, for which see: http://en.wikipedia.org/wiki/Delphi_technique
The Centrebet market is quite interesting, and there is probably money to be made even now by covering the central possibilities. Mind you I can’t loose! I put $50 on labor at $1.12 early on in the piece. If I get nervous I’ll cover this by putting $2 on labout to win less than 50 seats at $20. (You’d be mad to bet on a Coalition victory at $9 when you can get $20 covering a slightly larger part of the probability space!)
Oh, I’m sure that it works for what you’ve linked to, but how will Glug then rank the people predicting? On what basis?
To edward o: I think I see the logic in Glug’s approach. Asking you to nominate a probability is a way of asking how confident you are in your prediction. The more confidently you nominate a selection, and you are correct, the more points you get. So, both you and I could make the same selections in that we both pick the winners, but the one that assigns the greater probability – that is, the bravest – will beat the other. It is a form of tie breaker; otherwise, enough people would have exactly the same result if it was just a matter of a tick or a cross.
Now to the issue of relationship between the results for each seat. We know that each of the individual contests are not independent of each other; if there is a movement in the electorate to go one way – a swing – then all the seats, to some extent, will be affected. Yes, there will be some seats that will have issues that affect them only, but electorate “mood” will have a say in a general shift for or against the government that each seat has to contend with.
My advice, for what it is worth, to those having a bash at the Glug contest is to first make a prediction on the general swing, or lack of it, you think will occur. Then, as you go down the list, consider how your predicted swing will effect that seat, modifying your final decision by any electorate-specific issues you think should be considered. Then, I think the probability you assign should be a “brave” one: either in the nineties, or below 10. I think this will give you the best prospect of winning the competition. You have to be correct in your predictions, of course, but if you are you will gain maximum points. On this I differ from Peter Brent’s approach which was to assign probability scores the way he saw the actual probability occurring (http://www.mumble.com.au/). If Peter gets close, I think he will be beaten by someone with the same, or close to the same, predictions but who was more “confident”.
Whatever, this is good stuff. I hope Glug publish the whole results because I would love to see how close (or far away) I actually get.
Libs will gain Indooroopilly by a large margin.