'Equalising' crowd figures

[manicmagpie]

{This is going to be a long post, hopefully being the stats board there will be some who will find it interesting}

Hey guys, crowd stats has always been my little obsession, but of course with the uneven draw it's always been difficult to directly compare the yearly totals of each club.

During this off season I've been playing around with some methods of equalising the figures to try and remove the discrepancies of the draw. This method is the best I can come up with:

For each club, I've calculated the long term average (2000 onwards) against each opponent on each ground both as the home team and the away team. For each club I then take the average of those averages, splitting the opponents into derby and interstate clubs.

Taking the example of Collingwood as home team at the MCG, they average against Vic clubs:

Carlton        70,250
Essendon       76,662
Geelong        67,646
Hawthorn       60,668
Melbourne      46,622
North Melb     52,625
Richmond       59,642
St Kilda       53,698
W Bulldogs     49,614

The average of those averages is 59,715. I feel this represents the expected crowd for a Collingwood home game at the MCG against another Vic club. The same method for Collingwood at home at the MCG against non-Vic clubs gives a figure of 42,222.

Once all these calculations are done for every team on every applicable ground, you can calculate the expected attendance of the entire season for each club.

Using Essendon's fixture from last season as an example: they started with North at Docklands (Away), Port at Docklands (H) and GC at Carrara (A).

Therefore you find North's home crowd average at Docklands for derby games (33,854) + Port's away average at Docklands for interstate games (24,533) + GC's home average at Carrara for interstate games (15,779). This means that if every club had this draw they would average 74,166 across the 3 games. Essendon actually got 97,005 to these games, which is 131% of the average.

Extending this to all 2012 games for all clubs yields the following results:

               Actual     Expected    Total %
Collingwood    1,207,743  867,108    139.28%
Essendon       994,993    817,081    121.77%
Hawthorn       891,025    790,256    112.75%
Carlton        965,334    857,690    112.55%
Richmond       871,504    795,072    109.61%
West Coast     754,425    693,161    108.84%
Adelaide       659,884    639,499    103.19%
Geelong        760,616    744,665    102.14%
Sydney         595,825    609,557    97.75%
Fremantle      663,264    685,407    96.77%
Gold Coast     387,289    423,802    91.38%
St Kilda       695,296    763,040    91.12%
GWS            349,870    398,589    87.78%
North Melb     567,363    663,913    85.46%
Bris Lions     514,380    608,020    84.60%
Melbourne      618,341    759,398    81.43%
W Bulldogs     526,959    724,482    72.74%
Port Adel      453,641    637,665    71.14%

Problems
The main problem is the small sample size for several matchups. The most obvious difficulty being that there are two new clubs with new grounds that provide no meaningful long term average.

WC and Adelaide's percentages are possibly overstated as they share a ground with just one other team which consistently gets smaller crowds. Therefore all their home games look good by comparison.

There is also no allowance for timeslots. This is because:
a) even I can't be bothered with finding a separate average for each matchup at each ground for each different timeslot;
b) it would dilute the sample sizes to a useless degree; and
c) I think the effect of timeslots on crowd sizes is usually overstated. My next project might be to try and identify the effect, if any that the timeslot has. I have only had a look at Collingwood's figures by timeslot and there is no discernible pattern to them.

Hopefully I've explained the methodology clearly enough. Anyone who had the patience to read through all of this I would welcome your feedback.

 

[RogersResults]

{This is going to be a long post, hopefully being the stats board there will be some who will find it interesting}

Hey guys, crowd stats has always been my little obsession, but of course with the uneven draw it's always been difficult to directly compare the yearly totals of each club.

During this off season I've been playing around with some methods of equalising the figures to try and remove the discrepancies of the draw. This method is the best I can come up with:

For each club, I've calculated the long term average (2000 onwards) against each opponent on each ground both as the home team and the away team. For each club I then take the average of those averages, splitting the opponents into derby and interstate clubs.

Taking the example of Collingwood as home team at the MCG, they average against Vic clubs:

Carlton        70,250
Essendon      76,662
Geelong        67,646
Hawthorn      60,668
Melbourne      46,622
North Melb    52,625
Richmond      59,642
St Kilda      53,698
W Bulldogs    49,614

The average of those averages is 59,715. I feel this represents the expected crowd for a Collingwood home game at the MCG against another Vic club. The same method for Collingwood at home at the MCG against non-Vic clubs gives a figure of 42,222.

Once all these calculations are done for every team on every applicable ground, you can calculate the expected attendance of the entire season for each club.

Using Essendon's fixture from last season as an example: they started with North at Docklands (Away), Port at Docklands (H) and GC at Carrara (A).

Therefore you find North's home crowd average at Docklands for derby games (33,854) + Port's away average at Docklands for interstate games (24,533) + GC's home average at Carrara for interstate games (15,779). This means that if every club had this draw they would average 74,166 across the 3 games. Essendon actually got 97,005 to these games, which is 131% of the average.

Extending this to all 2012 games for all clubs yields the following results:

              Actual    Expected    Total %
Collingwood    1,207,743  867,108    139.28%
Essendon      994,993    817,081    121.77%
Hawthorn      891,025    790,256    112.75%
Carlton        965,334    857,690    112.55%
Richmond      871,504    795,072    109.61%
West Coast    754,425    693,161    108.84%
Adelaide      659,884    639,499    103.19%
Geelong        760,616    744,665    102.14%
Sydney        595,825    609,557    97.75%
Fremantle      663,264    685,407    96.77%
Gold Coast    387,289    423,802    91.38%
St Kilda      695,296    763,040    91.12%
GWS            349,870    398,589    87.78%
North Melb    567,363    663,913    85.46%
Bris Lions    514,380    608,020    84.60%
Melbourne      618,341    759,398    81.43%
W Bulldogs    526,959    724,482    72.74%
Port Adel      453,641    637,665    71.14%

Problems
The main problem is the small sample size for several matchups. The most obvious difficulty being that there are two new clubs with new grounds that provide no meaningful long term average.

WC and Adelaide's percentages are possibly overstated as they share a ground with just one other team which consistently gets smaller crowds. Therefore all their home games look good by comparison.

There is also no allowance for timeslots. This is because:
a) even I can't be bothered with finding a separate average for each matchup at each ground for each different timeslot;
b) it would dilute the sample sizes to a useless degree; and
c) I think the effect of timeslots on crowd sizes is usually overstated. My next project might be to try and identify the effect, if any that the timeslot has. I have only had a look at Collingwood's figures by timeslot and there is no discernible pattern to them.

Hopefully I've explained the methodology clearly enough. Anyone who had the patience to read through all of this I would welcome your feedback.

Its not clear what you are trying to demonstrate - there has always been correlation between performance and attendance.

If you are trying to demonstrate that the fixture determines attendance figures, then if you could discover the revenue distribution from matches this might mean something. (But its been a long time since even 'gate money' at matches has been published.)

As far as revenue distribution to the competing clubs is concerned, a crowd of 25,000 at Kardinia Park is not the same as 25,000 at Docklands which is not the same as 25,000 at Subiaco and so on. This has always been the case, regardless of whether the fixture was "even" or not. Raw attendance figures cannot give the full picture.

 

[manicmagpie]

Its not clear what you are trying to demonstrate - there has always been correlation between performance and attendance.

If you are trying to demonstrate that the fixture determines attendance figures, then if you could discover the revenue distribution from matches this might mean something. (But its been a long time since even 'gate money' at matches has been published.)

As far as revenue distribution to the competing clubs is concerned, a crowd of 25,000 at Kardinia Park is not the same as 25,000 at Docklands which is not the same as 25,000 at Subiaco and so on. This has always been the case, regardless of whether the fixture was "even" or not. Raw attendance figures cannot give the full picture.

Basically it's in response to the annual shitfights over crowd figures, which almost always includes arguments about the draw i.e. Team X got to play Collingwood twice while Team Y had to play in Launceston etc. I think these are valid arguments to have when considering that there is a gap of over 200,000 between the largest and smallest expected attendances of the Vic clubs.

So, boiling it down it's about trying to create a system to see "if every team were given that draw what would you expect the attendance to be" and measuring how far each team exceeded or fell short of that average. It's about attempting to remove the variable of the fixture from the crowd figures.

I understand that performance greatly affects crowd figures, this is not about adjusting for that aspect. Using this method over a longer period of time than one year would go some way to levelling out this factor, if it was a concern.

I'm also not making any statements about revenue, however I do think this exercise is useful in demonstrating how far behind some clubs start, courtesy of the fixture.

Basically what I'm after here is whether or not you think that this is a reasonable method for removing the variable of the fixture from the crowd figures. Are there any changes that you would make or any problems with the methodology?

 

[SJ]

I've always loved the idea of running crowd sizes against the following variables:
- Membership numbers (club A/club B, possibly split into type)
- Ground capacity
- Weather
- Time slot
- Television details (free to air live, free to air delayed, pay TV)
- Ladder position
- Recent form

And so on.

 

[RogersResults]

Basically it's in response to the annual shitfights over crowd figures, which almost always includes arguments about the draw i.e. Team X got to play Collingwood twice while Team Y had to play in Launceston etc. I think these are valid arguments to have when considering that there is a gap of over 200,000 between the largest and smallest expected attendances of the Vic clubs.

So, boiling it down it's about trying to create a system to see "if every team were given that draw what would you expect the attendance to be" and measuring how far each team exceeded or fell short of that average. It's about attempting to remove the variable of the fixture from the crowd figures.

I understand that performance greatly affects crowd figures, this is not about adjusting for that aspect. Using this method over a longer period of time than one year would go some way to levelling out this factor, if it was a concern.

I'm also not making any statements about revenue, however I do think this exercise is useful in demonstrating how far behind some clubs start, courtesy of the fixture.

Basically what I'm after here is whether or not you think that this is a reasonable method for removing the variable of the fixture from the crowd figures. Are there any changes that you would make or any problems with the methodology?

The problem is that you are not comparing like with like - each bum on seat/face in the crowd has a different value depending where it is.
For example:

"Skilled Stadium provides a significant financial windfall for home clubs, with Cats CEO Brian Cook yesterday stating a 20,000 crowd brings in about $600,000.
The advantages of moving to Geelong could see clubs earning close to $30 a head at Skilled, compared with about $10 a head at Etihad and a similar figure at the MCG."

http://www.geelongadvertiser.com.au/article/2010/06/09/180701_news.html

Unless you can discover each club's return per head at each venue plus factor in broadcasting revenue and the redistibution of revenue undertaken by the AFL's 'equalisation system' then working with just the raw attendance figures won't reveal anything of significance nor provide information to make the the competition more 'fair' or 'even'. Collingwood playing Nth Melbourne and the Western Bulldogs twice in a season instead of Carlton and Essendon could actually mean less revenue for North and the Bulldogs - just the attendance figures won't answer that.

All in all, a huge task to attempt to demonstate that the League has got it wrong.

 

[RogersResults]

Maybe some useful information here:
http://footybusiness.wordpress.com/

 

[RogersResults]

I've always loved the idea of running crowd sizes against the following variables:
- Membership numbers (club A/club B, possibly split into type)
- Ground capacity
- Weather
- Time slot
- Television details (free to air live, free to air delayed, pay TV)
- Ladder position
- Recent form

And so on.

R16 2005 Coll (13th) v Ess (14th) MCG Fri. night - 52,507 (both 12 points and 33% outside the top 8.)
R19 2008 WB (2nd) v NM (5th) Dklnd Sun. day - 31,957 (Nth out of the top 4 on percentage only.)

(R15 1999 NM (3rd) v WB (4th) MCG Sun. day - 44,683.)

Draw your own conclusions if any.

 

[manicmagpie]

The problem is that you are not comparing like with like - each bum on seat/face in the crowd has a different value depending where it is.
For example:

"Skilled Stadium provides a significant financial windfall for home clubs, with Cats CEO Brian Cook yesterday stating a 20,000 crowd brings in about $600,000.
The advantages of moving to Geelong could see clubs earning close to $30 a head at Skilled, compared with about $10 a head at Etihad and a similar figure at the MCG."

http://www.geelongadvertiser.com.au/article/2010/06/09/180701_news.html

Unless you can discover each club's return per head at each venue plus factor in broadcasting revenue and the redistibution of revenue undertaken by the AFL's 'equalisation system' then working with just the raw attendance figures won't reveal anything of significance nor provide information to make the the competition more 'fair' or 'even'. Collingwood playing Nth Melbourne and the Western Bulldogs twice in a season instead of Carlton and Essendon could actually mean less revenue for North and the Bulldogs - just the attendance figures won't answer that.

All in all, a huge task to attempt to demonstate that the League has got it wrong.

We're talking at cross purposes. It's my fault for using the word 'equalising' in the thread title which you have linked to the AFL's 'equalisation' policies. It's nothing to do with that (although there is scope to go in that direction, but as you say, it must be coupled with stadium revenue data amongst other things).

It's simply about my nerdish fascination with the crowd figures and that how comparing how clubs fared is massively affected by the uneven fixture. It's about trying to settle some of the d*ick measuring arguments of the type found in the annual attendances thread on the main board, of which a recurring theme is 'we would have had bigger crowds than you except you got a better draw'.

 

[RogersResults]

We're talking at cross purposes. It's my fault for using the word 'equalising' in the thread title which you have linked to the AFL's 'equalisation' policies. It's nothing to do with that (although there is scope to go in that direction, but as you say, it must be coupled with stadium revenue data amongst other things).

It's simply about my nerdish fascination with the crowd figures and that how comparing how clubs fared is massively affected by the uneven fixture. It's about trying to settle some of the d*ick measuring arguments of the type found in the annual attendances thread on the main board, of which a recurring theme is 'we would have had bigger crowds than you except you got a better draw'.

I see that the issue is similiar to the nonsense thread about memberships that ignores REVENUE - for instance Geelong in 2011 had significantly more membership revenue than Hawthorn despite having 10,000 or so fewer 'members'.

I'm sure no thinking Geelong supporter cares about 20,000 attendances in Geelong when the financial return to the club is twice or more than a full-house returns to a 'home' team at Docklands. The same applies to the interstate teams and Hawthorn in Tasmania.

The fact is some clubs attract significantly more spectators than do other clubs regardless. The League would argue that return matches between the highest drawing clubs in Melbourne along with 'derbies' in other states, benefits all clubs by the extra revenue generated.

The only way to get this across to the 'd*ck measurers' is to somehow get them to grasp that its all about revenue v costs and if a fixture at 'x' venue attracts two thirds of a crowd than does another fixture at 'y' venue but returns twice the amount of revenue to the club, then that's what they should be crowing about not about the meaningless number of bodies at the match. I doubt your 'equalising' formula would convince them of anything even if they actually began to understand it.

 

[SJ]

R16 2005 Coll (13th) v Ess (14th) MCG Fri. night - 52,507 (both 12 points and 33% outside the top 8.)
R19 2008 WB (2nd) v NM (5th) Dklnd Sun. day - 31,957 (Nth out of the top 4 on percentage only.)

(R15 1999 NM (3rd) v WB (4th) MCG Sun. day - 44,683.)

Draw your own conclusions if any.

That's where I assume the membership numbers will kick in.