Poll Way Aggregate: ALP 50.8%, L/NP 49.2%

Well, August has rolled around so it’s about time for a new Poll Way Aggregate. Since our last look at political polling on 16 July we’ve had a brace of new polls come in while fights over asylum seekers, cars, the budget and carbon pricing have kicked off or continued.

As always, polls older than one calendar month have been excluded, as have polls that fall outside the margin of error of subsequent polls from that same House. The polls used in the most recent Aggregate are below, including a total of 20,116 respondents:

HOUSE

DATE

ALP   (2PP)

L/NP (2PP)

Newspoll

5-7/7/2013

50%

50%

Nielsen

11-13/7/2013

50%

50%

Essential

11-14/7/2013

48%

52%

Morgan (Multi)

12-14/7/2013

52.5%

47.5%

ReachTEL

18/07/2013

49%

51%

Morgan (Multi)

19-21/7/2013

52.5%

47.5%

Newspoll

19-21/7/2013

48%

52%

Galaxy

23-25/7/2013

50%

50%

Essential

25-28/7/2013

49%

51%

Morgan (Multi)

26-28/7/2013

52%

48%

Since the last Aggregate, the ALP has had a slight uptick of less than 1 per cent, which really suggests that there has been no real change. The narrative in the media over the last two weeks regarding polling has largely been that Labor’s momentum has stalled, and while that’s one way to read these numbers, the other might be that the polls are actually accurately measuring what the electorate’s views are. Under Gillard, Labor was unpopular while under Rudd, Labor is in with a chance – there has been little change because the views of people have been constant.

In the last Aggregate we discussed the effects of the Morgan (Multi) poll, and how it might be over-inflating Labor’s figures. Once again Morgan has been consistently coming in higher than the other polls as seen in both the two-party preferred and primary vote tables:

HOUSE

DATE

ALP 1°

L/NP 1°

GRN 1°

‘Other’ 1°

Newspoll

5-7/7/2013

38%

42%

9%

11%

Nielsen

11-13/7/2013

39%

44%

9%

8%

Essential

11-14/7/2013

39%

46%

7%

8%

Morgan (Multi)

12-14/7/2013

42%

41%

7%

10%

ReachTEL

18/07/2013

39.3%

45.4%

8.3%

7%

Morgan (Multi)

19-21/7/2013

41.5%

41%

9%

8.5%

Newspoll

19-21/7/2013

37%

45%

10%

8%

Galaxy

23-25/7/2013

40%

44%

9%

7%

Essential

25-28/7/2013

39%

44%

9%

8%

Morgan (Multi)

26-28/7/2013

38.5%

41.5%

10.5%

9.5%

The thing is, though, despite being consistently higher, Morgan is consistent. As are all the polls with their usual House biases. Other analyses of polls adjust for these perceived biases while we at the Poll Way avoid doing so for fear of introducing our own bias. As the numbers come in, we’ll just put them up. That being said, a less than 2 point spread in the 2PP scores is consistent with most of the polls as well as analyses done elsewhere. The only story that anyone can anyone can confidently tell is that it’s close, very close. The Poll Way’s own numbers, however, do detect a small swing to the ALP at both a two-party preferred and primary vote level with primary vote scores coming in at: ALP 39.8%, L/NP 42.8%, GRN 8.8% and ‘Other’ 8.7%.

The primary vote shows almost 1 for 1 stability for the Coalition and Green votes from the last aggregate, with a movement of 1% to the ALP coming almost exclusively from the ‘Other’ column - probably as disenfranchised non-lefty voters come back to Labor after flirting with third parties and independents.

Anecdotally, there are stories of many traditional Labor voters who were turned off by Julia Gillard but who wouldn’t vote Liberal or Green and who were attracted to start-ups like Katter et al. They clearly exist and they are coming back to Labor.

Round 17 Review and Richmond v Fremantle

Well, that didn’t work!

Despite Fremantle having a much higher average Champion Data score for the team they took to the MCG to take on Richmond, the result didn’t go their way with Richmond getting up by 27 points.

Interestingly once you add to the team averages the Champion Data scores for each player for this game, Richmond’s score goes up by so much, and Fremantle’s down by so much, that Richmond actually ends up ahead (Richmond 70.1 and Fremantle 68.8). This isn’t really useful for our purposes as there’s no point in using data from future events to predict past performance. However, maybe Champion Data’s player rankings are measuring something they should be - they’re just not necessarily good as a predictive measure (at least not on their own…)

It should also be noted further to the last post that Collingwood achieved higher statistics in nearly all measures in their match despite losing to Gold Coast, while Sydney was beaten in many metrics by West Coast yet they cruised to an easy victory.

The pictures of a very efficient Sydney and a woefully inefficient Collingwood are becoming harder and harder to dismiss.

Champion Data Averages and Round 17: Richmond v Fremantle

Champion Data has a contract with the AFL to provide statistics for every player and team for every match played during the Home and Away and Finals series, with the data being provided to the AFL and published publicly here. This data is no doubt used for a number of purposes, one of them being the creation of player ratings.

These player ratings interest the Poll Way, as they purport to describes the relative performances of players in any given game, and to build a score for players based on their performances over the past 40 rounds. If these ratings are accurate over the course of a season, then it might be possible to build a model to predict the outcomes of matches come finals time. A full explanation of how player ratings work can be found here.

For the purposes of a very simple initial investigation of this idea, for this week the Poll Way is going to pretend that Round 17 is a finals round. As of the start of Round 17, the top 8 teams on the ladder are Collingwood, Essendon, Fremantle, Geelong, Hawthorn, Port Adelaide, Richmond and Sydney. Only matches between these teams to date during the 2013 season will be included to develop team ratings. Matches with the bottom 10 teams won’t be used, as the Poll Way takes the view that there’s no point in including Jobe Watson’s performance against, say, Melbourne, as it will have little to no bearing on how Essendon will do against a true finals contender.

There have been 25 games played between these 8 teams so far in 2013. The Poll Way has taken the player ratings for each player in each match and averaged them to provide a team score for every match. Theoretically, the team with the higher rating should be the team that won. The table below shows that this isn’t always the case.

HOME   TEAM

AWAY   TEAM

WINNER

HIGHEST   MATCH RANKING

Hawthorn

Geelong

Geelong

Hawthorn

Collingwood

Hawthorn

Hawthorn

Hawthorn

Fremantle

Essendon

Essendon

Fremantle

Richmond

Collingwood

Collingwood

Collingwood

Hawthorn

Fremantle

Hawthorn

Hawthorn

Sydney

Geelong

Sydney

Sydney

Essendon

Collingwood

Essendon

Essendon

Fremantle

Richmond

Fremantle

Fremantle

Richmond

Geelong

Geelong

Geelong

Fremantle

Collingwood

Fremantle

Collingwood

Geelong

Essendon

Geelong

Essendon

Hawthorn

Sydney

Hawthorn

Hawthorn

Port Adelaide

Richmond

Richmond

Richmond

Collingwood

Geelong

Collingwood

Collingwood

Sydney

Fremantle

Sydney

Sydney

Collingwood

Sydney

Sydney

Sydney

Richmond

Essendon

Essendon

Essendon

Port Adelaide

Geelong

Geelong

Geelong

Sydney

Essendon

Sydney

Sydney

Port Adelaide

Sydney

Port Adelaide

Port Adelaide

Port Adelaide

Collingwood

Port Adelaide

Port Adelaide

Geelong

Fremantle

Geelong

Geelong

Essendon

Port Adelaide

Essendon

Essendon

Geelong

Hawthorn

Geelong

Geelong

Port Adelaide

Hawthorn

Hawthorn

Hawthorn

The above table is a bit difficult to navigate, but in short it shows that generally high Champion Data match rankings accord with match wins. There are 4 occasions where it doesn’t work out like that though, namely Hawthorn v Geelong, Fremantle v Essendon, Fremantle v Collingwood and Geelong v Essendon.

There is another way to use the collected data to create a metric of team performance, which is to average the match rankings of individual players over the course of the relevant matches and then average the scores of the 22 players who were selected for a given match. These results are given below.

HOME   TEAM

AWAY   TEAM

WINNER

HIGHEST   AVERAGE RANKING

Hawthorn

Geelong

Geelong

Hawthorn

Collingwood

Hawthorn

Hawthorn

Hawthorn

Fremantle

Essendon

Essendon

Essendon

Richmond

Collingwood

Collingwood

Collingwood

Hawthorn

Fremantle

Hawthorn

Hawthorn

Sydney

Geelong

Sydney

Sydney

Essendon

Collingwood

Essendon

Essendon

Fremantle

Richmond

Fremantle

Fremantle

Richmond

Geelong

Geelong

Geelong

Fremantle

Collingwood

Fremantle

Collingwood

Geelong

Essendon

Geelong

Geelong

Hawthorn

Sydney

Hawthorn

Hawthorn

Port Adelaide

Richmond

Richmond

Richmond

Collingwood

Geelong

Collingwood

Geelong

Sydney

Fremantle

Sydney

Fremantle

Collingwood

Sydney

Sydney

Collingwood

Richmond

Essendon

Essendon

Essendon

Port Adelaide

Geelong

Geelong

Geelong

Sydney

Essendon

Sydney

Essendon

Port Adelaide

Sydney

Port Adelaide

Sydney

Port Adelaide

Collingwood

Port Adelaide

Collingwood

Geelong

Fremantle

Geelong

Geelong

Essendon

Port Adelaide

Essendon

Essendon

Geelong

Hawthorn

Geelong

Hawthorn

Port Adelaide

Hawthorn

Hawthorn

Hawthorn

Oh dear… All of a sudden there are 9 games where the winner is not the team with the highest ranking.

However, each of these 9 games can be placed in one of three different categories:
1. Matches with Collingwood
2. Matches with Sydney
3. Matches between Geelong and Hawthorn

Where the results of matches with Collingwood and Sydney are inconsistent with the average rankings, Collingwood has a tendency to lose matches it shouldn’t, while Sydney has a tendency to win matches it shouldn’t.

While the Poll Way doesn’t necessarily have an explanation for this, it does suggest that Collingwood and Sydney have considerable inconsistencies between their ability to get hold of the ball, and their ability to score. Collingwood have an excellent midfield capable of cutting through most other teams, yet their forwards aren’t necessarily able to convert. Conversely, Sydney appears to be able to score efficiently, perhaps due in part to their now famous ‘slingshot’ off the half-back line which can deliver coast to coast goals while racking up few stats.

As for games between Geelong and Hawthorn, well, Geelong did promise that they’d never lose another game to Hawthorn again. The Kennett curse is alive and well (but then again, we already knew that).

Now, all this numberin’ wouldn’t be any fun unless we tried to predict something so we can get it wrong or right. Given that the Poll Way is pretending that round 17 is the finals, and that there is only one match between top 8 sides this week, we will look at that match: Richmond v Fremantle.

Using the method described for the second table above, Fremantle has an average player rating of 69.7, while Richmond’s score is 67.9. Two players who will be in this match have not played a game against another top 8 side this year so far, namely Aaron Sandilands for Fremantle and Matthew Arnot for Richmond. Both of these players have been assigned a score of 0 by the Poll Way for the purposes of this analysis. This is of course inaccurate, however with all due respect to Mr Arnot, Sandilands will probably have a greater positive impact on the game. Sandilands’ true score is likely greater than Arnot’s, so the fact that they have both been assigned a score of 0 and Fremantle still has a higher overall rating than Richmond means that there is no overall change to the two teams’ relative rankings. By these numbers, despite Pavlich and Ballantyne being significant outs for Fremantle, they should still get up over Richmond.

Then again, anything can happen on the day.

Poll Way Aggregate: ALP 50.2%, L/NP 49.8%

Welcome to the Poll Way’s new home on the internet. We’ve finally moved into our new digs and we have to say, conducting our business on these new-fangled intarwebs is quite the excitement.

With the release of this week’s Nielsen poll, something in the order of 21,000 Australians have now been polled nationally since the election of Kevin Rudd as Labor leader by caucus on 26 June. Nielsen is also the last of the 5 major polling houses in Australia to release results since Kevin Rudd’s return as leader, which the Poll Way is marking by conducting it’s first Aggregate since the new regime.

As always, the Aggregate represents a weighted average of all polls, excluding every second Essential poll (to avoid double counting samples included in their rolling two week averages), any polls that fall outside of the margin of error of that house’s subsequent polls, and any poll older that one calendar month after it was completed. For the purposes of the Aggregate, Morgan polls using different techniques are treated as different polls. These criteria give us a sample size of 15,524 electors. Two party preferred results are given below.

HOUSE

DATE

ALP (2PP)

L/NP (2PP)

Morgan (SMS)

26/6/2013

49.5%

50.5%

ReachTEL

27/6/2013

48%

52%

Galaxy

27-28/6/2013

49%

51%

Newspoll

28-30/6/2013

49%

51%

Essential

4-7/7/2013

48%

52%

Morgan (Multi)

5-7/7/2013

54.5%

45.5%

Newspoll

5-7/7/2013

50%

50%

Nielsen

11-13/7/2013

50%

50%

The first thing you’ll notice about the above table is that despite the excitement created in the community around Kevin Rudd’s return, most of the polls since 26 June have come in in favour of the Coalition, which would give us a straight two party preferred average of ALP 49.75% and L/NP 50.25%. However, the Morgan (Multi) poll in the above table comes in heavily for Labor and, with a sample size of over 3,500 respondents, tilts the Poll Way Aggregate in favour of them, ALP 50.2%, L/NP 49.2% – essentially swapping the result of a straight average.

The Poll Way Aggregate has also been calculated for primary vote values.

HOUSE

DATE

ALP (Primary)

L/NP (Primary)

GRN (Primary)

‘Other’ (1°)

Morgan (SMS)

26/6/2013

38%

43%

8.5%

10.5%

ReachTEL

27/6/2013

38.3%

45.1%

8.7%

7.8%

Galaxy

27-28/6/2013

38%

44%

10%

8%

Newspoll

28-30/6/2013

35%

43%

11%

11%

Essential

4-7/7/2013

38%

46%

8%

7%

Morgan (Multi)

5-7/7/2013

41.5%

39.5%

8.5%

10.5%

Newspoll

5-7//2013

38%

42%

9%

11%

Nielsen

11-13/7/2013

39%

44%

9%

8%

Poll watchers will see that the most striking feature of the primary vote is the dramatic recovery of Labor since Kevin Rudd’s reinstallment, shifting Labor out of the high 20s and into the high 30s – a base from which they can contest an election. Having said that, though, the Coalition primary vote looks relatively secure. There has been an initial drop since June 26, however it is not by as much as Labor’s has increased and has remained stable since. This makes the Morgan (Multi) results appear even more curious, as they have the Coalition dropping below 40%, and even have Labor ahead on primary. This seems an unlikely proposition given both historical results and other polling data coming through. We’ll see what changes flow through as more data comes in over time.

The Poll Way has not had a good look at the relative values of the Green and ‘Other’ primary votes over time, however a cursory glance suggests that the Green vote is remaining stable and the ‘Other’ vote is declining slightly. This is not unexpected given the Greens supporter base and the rebuilding of Labor’s primary vote – those votes have to come from somewhere. Bob Katter and Clive Palmer might have to hold their horses, or hats, or whatever, for a while longer yet.

Having said all that, the Poll Way Aggregate primary votes are as follows: ALP 38.1%, L/NP 43%, GRN 8.8%, Other 9.2%.