How good are the Individual Seat Odds from the Bookies at predicting the winner of the individual seat?

Below are some results from the 2007 election. (Click for a larger image)

These show some of the seats with the winner, two-party preferred vote
after the election and before and the swing away from ALP. They
show the Sportingbet probability of ALP winning the seat as offered on
Nov 23rd, election eve. Thanks Simon Jackman for keeping that
data.

We also see the winner of the seat according to three different prediction
methods.

First is by using the uniform national swing reported in the Newspoll
of Nov 23.

Next is the winner according to Sportingbet. (Sportingbet didn't
give odds for Hinkler and Kennedy like many seats that
were assumed to be very safe. So we just assume that
Sportingbet was assuming that the current holder would
win.)

Last is the winner according to state by state uniform swings
as predicted by Newspoll on Nov 23rd.

There are 26 seats listed. All three methods agreed in their predictions
on all the other 124 seats and they were all correct
in those predictions. These are the seats where at least one method
got the winner wrong.





















So what was the best way to predict the winner of a seat?

Here are the results. Recall that the ALP won 83 of the
150 seats. (They won 110 of the seats not listed here).

Method 1: (use the latest Newspoll and assume a uniform
national swing)
You get 134 predictions correct and you think the ALP will
win 75 seats.

Method 2: (use sportingbet odds for each set)
You get 138 predictions correct and you think the ALP will
win 81 seats.

Method 3: (Use the latest Newspoll state by state swings,
with the national swing for NT, ACT and Tas)
You get 132 predictions correct and you think that the ALP
will win 83 seats.

So, the most accurate predictor of individual seat winners
was to use the bookies' odds.

You can do some more work and find out the expected
number of seats won by each party but we'll come
back to that later.