Seat betting market update

Turning individual seat odds into probabilities is vexed. For the decimal odds used by online bookmakers, the maths is simple:

$$probability = \frac{1}{odds}$$

The problem starts with the bookmaker's over-round. If we sum the odds for every possible outcome, the probability exceeds 1. For the individual seat prediction markets, the over-round can exceed 1 by as much as 30%.

Let's look at the seat of Hunter in NSW. This morning, Sportsbet had Labor on odds of $1.01, which using the above formula gives Labor a win probability of 99 per cent in Hunter. So far so good.

But Sportsbet also has odds for the Coalition ($10.00), One Nation ($12.00) and the United Australia Party ($11.00). These equate to a 10 per cent probability for the Coalition, an 8.33 per cent probability for One Nation and a 9.09 per cent probability for the United Australia Party. The bookmaker's over-round is 126.43 per cent.

If we do a simple normalisation of these probabilities to correct for the over-round, we have Labor on a 78.3 per cent chance of winning. The Coalition has a 7.9 per cent chance of winning. One Nation has a 6.6 per cent chance, and the United Australia Party has a 7.2 per cent chance. Together, these add to 100 per cent.

However, the normalised probabilities understate the probability for Labor and over-state the probability for the other parties. According to the 2019 election pendulum, Hunter is a safe Labor seat. There is no way that Labor has a one in five chance of losing Hunter at the 2019 election. I suspect this probability outcome comes about from having such a large bookmakers' over-round, and from a phenomenon known as the long-shot bias.

In this week's analysis, I have taken a more aggressive approach to correcting for the longshot bias and the large over-rounds.

• For odds between $1.01 and $1.02 (bookmaker raw probabilities between 98 and 99 per cent), I have treated the seat as having a probability of 100 per cent for the favourite.  
• For odds between $1.03 and $1.50 (bookmaker raw probabilities between 66.7 and 97 per cent), I have done all of the normalisation for the bookmaker's over-round in terms of the non-favourite parties.  
• For odds in excess of $10, I have simply ignored them.

As a result, this week's analysis does not compare easily with last week's analysis of seat betting markets. In last week's analysis, I only excluded odds over $15.00.

Well, with that out of the way, let's look at the results. We will start with the summary of individual seat outcomes (ordered high to low in respect of the Coalition's probability of winning).

In terms of an election outcome, this translates as follows.

And the time-series charts for individual seats are as follows. You will note that I could not collect odds for a couple of Tasmanian seats on a couple of days this week. Also note: I usually collect these odds auto-magically in the morning while I am having breakfast (I have written a fairly simple web-scrapper for this purpose).