Don’t ask me why, but I found myself looking at some of the factors that have affected Palace’s home attendances over the last 20 years or so. I thought the results may be of interest to some so I’ve summarised them below. If you don’t like this nerdy stuff, please save yourself some time and look elsewhere.
The data I’ve used is since the 1994-95 season i.e. after the demolition of the Holmesdale Road terrace. I’ve focused on the league for simplicity. Some of the results may be stating the obvious to a degree, but it at least puts some numbers on it. Trying to draw inferences about what this would mean for Palace in a bigger ground is beyond the scope of what I have estimated and would be wildly speculative. But I’ll do it a bit anyway. I could do a quick post about my methodology for anyone that gives a stuff about this sort of thing. Suffice to say, I’ve used a Tobit censored regression approach but I haven’t spent a great deal of time on diagnostics so I’m not pretending this has been a rigorous exercise. Nonetheless, the results seem to make some intuitive sense.
What is included
The ‘quality’ of Palace that season; the opposition ‘quality’; the day the match was played; whether the match was on TV (post circa 2000); what league we were in; whether we were playing a London club and the time of year. I have also controlled for playoff matches and certain seasons that distort the results.
What is not included
96/97 season – data for this season was missing from the sources I looked at for some reason.
Ticket prices – not sure if a decent archive of ticket prices would exist? Might be difficult to isolate any effect on a season by season basis but there may be some particular matches that were affected by price promotions etc (see the results for any that stand out)
Club specific effects – you’ll see in the results, other than making an allowance for London clubs, I don’t try to specifically control for specific clubs. Some stick out like a sore thumb in the residuals but that is interesting in itself.
Anything else – I’m limited to data that are easily available. I’ve tried not to be too subjective and enter in things like “big games” but if there is anything else people can think of, let me know.
Results/Observations (stating the obvious but these are statistical estimates – with uncertainty)- Palace attracts glory hunters. But not that much. For each place higher Palace are in the league, we attract 165 extra people (other things being equal – this is an important point and true of all numbers).
- Ok, maybe quite a bit. Being in the Premier League alone adds 9,000 to our attendances.
- We don’t care too much how good our opponents are. For every place higher our opponent is in the league, we add 95 people.
- Weekday matches are worse attended. But not by that much: 1200 lower in fact.
- Perhaps surprisingly, being on TV does not have a statistically significant effect on our attendances.
- We have slightly bigger attendances against London clubs. An average of 540 higher although Wimbledon probably distort this downwards a bit.
- After Christmas our attendances rise by around 700 on average.
- Our mid-late 1990s attendances were fairly poor. I’m fully aware that in 94-95 we were down a stand but my methodology should account for this. I’ve controlled specifically for all seasons from 94/95 to 97/98 and they were over 2000 lower on average.
- Our attendances have gone up a notch. Now this is stating the obvious because we mostly sell out now. But in 15/16 we were 2,000 above what would be explained by league placings etc.
- The model would suggest that we would currently be peaking at around 30,000 attendances if we removed capacity constraints. This is the wildly speculative part as we can’t observe anything. And that may be undercooking it a bit for big games. But the key is that league-position-specific effects would not be sufficient to increase our attendances very much. We would need to keep replicating the level effect we saw in 15/16 to regularly fill a big stadium. Would a few more seasons in the Premier League build on this? Would a new ground in itself attract more people?
Finally, you may be interested to see the biggest model ‘misses’ (the largest residuals in a technical sense).
Here are the top 10 ‘low’ attendances:
Team / Date / Result / Score / Attendance / Predicted / Gap
1 Wimbledon / 09/02/1998 / Lost / 0-3 / 14,410 / 20,742 / -6,332 /
2 Sheffield Wednesday / 14/03/1995 / Won / 2-1 / 10,422 / 14,516 / -4,094
3 Cardiff City / 22/09/2012 / Won / 3-2 / 12,757 / 17,859 / -5,102
4 Wimbledon / 17/09/1994 / Drawn / 0-0 / 12,100 / 15,567 / -3,467
5 Coventry City / 11/02/1995 / Lost / 0-2 / 11,871 / 15,238 / -3,367
6 Charlton Athletic / 05/12/2004 / Lost / 0-1 / 20,705 / 25,528 / -4,823
7 Sheffield Wednesday / 10/05/1998 / Won / 1-0 / 16,876 / 21,383 / -4,507
8 Sheffield United / 22/09/2007 / Won / 3-2 / 14,131 / 18,358 / -4,227
9 Cardiff City / 04/03/2008 / Drawn / 0-0 / 13,446 / 17,530 / -4,084
10 Birmingham City / 24/02/2009 / Drawn / 0-0 / 12,847 / 16,903 / -4,056
(*the size of the ‘gap’ is not in perfect order as there were changes in effective capacity over time – the model estimates proportion of capacity, not capacity directly)
I’m assuming that 6,332 individuals had premonitions of Carl Leaburn scoring 2 goals against us and so wisely decided to stay away from #1. Some of these show that our Premier League attendances were not always great. Perhaps not the draw in the mid-90s?
On to the top 10 ‘high’ attendances:
Team / Date / Result / Score / Attendance / Predicted / Gap
1 Brighton / 18/10/2005 / Lost / 0-1 / 22,400 / 15,634 / 6,766
2 Brighton / 26/10/2002 / Won / 5-0 / 21,796 / 15,861 / 5,935
3 West Ham United / 12/04/2004 / Won / 1-0 / 23,977 / 18,056 / 5,921
4 Leeds United / 25/04/2011 / Won / 1-0 / 20,142 / 14,780 / 5,362
5 Peterborough United / 04/05/2013 / Won / 3-2 / 22,154 / 16,900 / 5,254
6 Crewe Alexandra / 17/11/2001 / Won / 4-1 / 21,802 / 16,532 / 5,270
7 Burnley / 04/05/2008 / Won / 5-0 / 23,950 / 18,764 / 5,186
8 Reading / 29/10/2011 / Drawn / 0-0 / 21,002 / 16,206 / 4,796
9 Newcastle United / 29/11/1997 / Lost / 1-2 / 26,085 / 20,944 / 5,141
10 Sunderland / 05/04/1999 / Drawn / 1-1 / 22,096 / 17,133 / 4,963
The first two need no explanation. A “Br*ghton effect” of around 6,000. Though the 12/13 attendances against them were not that spectacular – perhaps we got sick of them again. Most of the others are important end of season games. Crewe and Reading are conspicuous ones though – was anything going on there with tickets? Also interesting how well we performed in these games – with a very notable aberration.
Overall, the model explains a little over 70% of our variation in attendances. I could add a few things in future if anyone has any ideas. I hope this has been of interest.
The data I’ve used is since the 1994-95 season i.e. after the demolition of the Holmesdale Road terrace. I’ve focused on the league for simplicity. Some of the results may be stating the obvious to a degree, but it at least puts some numbers on it. Trying to draw inferences about what this would mean for Palace in a bigger ground is beyond the scope of what I have estimated and would be wildly speculative. But I’ll do it a bit anyway. I could do a quick post about my methodology for anyone that gives a stuff about this sort of thing. Suffice to say, I’ve used a Tobit censored regression approach but I haven’t spent a great deal of time on diagnostics so I’m not pretending this has been a rigorous exercise. Nonetheless, the results seem to make some intuitive sense.
What is included
The ‘quality’ of Palace that season; the opposition ‘quality’; the day the match was played; whether the match was on TV (post circa 2000); what league we were in; whether we were playing a London club and the time of year. I have also controlled for playoff matches and certain seasons that distort the results.
What is not included
96/97 season – data for this season was missing from the sources I looked at for some reason.
Ticket prices – not sure if a decent archive of ticket prices would exist? Might be difficult to isolate any effect on a season by season basis but there may be some particular matches that were affected by price promotions etc (see the results for any that stand out)
Club specific effects – you’ll see in the results, other than making an allowance for London clubs, I don’t try to specifically control for specific clubs. Some stick out like a sore thumb in the residuals but that is interesting in itself.
Anything else – I’m limited to data that are easily available. I’ve tried not to be too subjective and enter in things like “big games” but if there is anything else people can think of, let me know.
Results/Observations (stating the obvious but these are statistical estimates – with uncertainty)- Palace attracts glory hunters. But not that much. For each place higher Palace are in the league, we attract 165 extra people (other things being equal – this is an important point and true of all numbers).
- Ok, maybe quite a bit. Being in the Premier League alone adds 9,000 to our attendances.
- We don’t care too much how good our opponents are. For every place higher our opponent is in the league, we add 95 people.
- Weekday matches are worse attended. But not by that much: 1200 lower in fact.
- Perhaps surprisingly, being on TV does not have a statistically significant effect on our attendances.
- We have slightly bigger attendances against London clubs. An average of 540 higher although Wimbledon probably distort this downwards a bit.
- After Christmas our attendances rise by around 700 on average.
- Our mid-late 1990s attendances were fairly poor. I’m fully aware that in 94-95 we were down a stand but my methodology should account for this. I’ve controlled specifically for all seasons from 94/95 to 97/98 and they were over 2000 lower on average.
- Our attendances have gone up a notch. Now this is stating the obvious because we mostly sell out now. But in 15/16 we were 2,000 above what would be explained by league placings etc.
- The model would suggest that we would currently be peaking at around 30,000 attendances if we removed capacity constraints. This is the wildly speculative part as we can’t observe anything. And that may be undercooking it a bit for big games. But the key is that league-position-specific effects would not be sufficient to increase our attendances very much. We would need to keep replicating the level effect we saw in 15/16 to regularly fill a big stadium. Would a few more seasons in the Premier League build on this? Would a new ground in itself attract more people?
Finally, you may be interested to see the biggest model ‘misses’ (the largest residuals in a technical sense).
Here are the top 10 ‘low’ attendances:
Team / Date / Result / Score / Attendance / Predicted / Gap
1 Wimbledon / 09/02/1998 / Lost / 0-3 / 14,410 / 20,742 / -6,332 /
2 Sheffield Wednesday / 14/03/1995 / Won / 2-1 / 10,422 / 14,516 / -4,094
3 Cardiff City / 22/09/2012 / Won / 3-2 / 12,757 / 17,859 / -5,102
4 Wimbledon / 17/09/1994 / Drawn / 0-0 / 12,100 / 15,567 / -3,467
5 Coventry City / 11/02/1995 / Lost / 0-2 / 11,871 / 15,238 / -3,367
6 Charlton Athletic / 05/12/2004 / Lost / 0-1 / 20,705 / 25,528 / -4,823
7 Sheffield Wednesday / 10/05/1998 / Won / 1-0 / 16,876 / 21,383 / -4,507
8 Sheffield United / 22/09/2007 / Won / 3-2 / 14,131 / 18,358 / -4,227
9 Cardiff City / 04/03/2008 / Drawn / 0-0 / 13,446 / 17,530 / -4,084
10 Birmingham City / 24/02/2009 / Drawn / 0-0 / 12,847 / 16,903 / -4,056
(*the size of the ‘gap’ is not in perfect order as there were changes in effective capacity over time – the model estimates proportion of capacity, not capacity directly)
I’m assuming that 6,332 individuals had premonitions of Carl Leaburn scoring 2 goals against us and so wisely decided to stay away from #1. Some of these show that our Premier League attendances were not always great. Perhaps not the draw in the mid-90s?
On to the top 10 ‘high’ attendances:
Team / Date / Result / Score / Attendance / Predicted / Gap
1 Brighton / 18/10/2005 / Lost / 0-1 / 22,400 / 15,634 / 6,766
2 Brighton / 26/10/2002 / Won / 5-0 / 21,796 / 15,861 / 5,935
3 West Ham United / 12/04/2004 / Won / 1-0 / 23,977 / 18,056 / 5,921
4 Leeds United / 25/04/2011 / Won / 1-0 / 20,142 / 14,780 / 5,362
5 Peterborough United / 04/05/2013 / Won / 3-2 / 22,154 / 16,900 / 5,254
6 Crewe Alexandra / 17/11/2001 / Won / 4-1 / 21,802 / 16,532 / 5,270
7 Burnley / 04/05/2008 / Won / 5-0 / 23,950 / 18,764 / 5,186
8 Reading / 29/10/2011 / Drawn / 0-0 / 21,002 / 16,206 / 4,796
9 Newcastle United / 29/11/1997 / Lost / 1-2 / 26,085 / 20,944 / 5,141
10 Sunderland / 05/04/1999 / Drawn / 1-1 / 22,096 / 17,133 / 4,963
The first two need no explanation. A “Br*ghton effect” of around 6,000. Though the 12/13 attendances against them were not that spectacular – perhaps we got sick of them again. Most of the others are important end of season games. Crewe and Reading are conspicuous ones though – was anything going on there with tickets? Also interesting how well we performed in these games – with a very notable aberration.
Overall, the model explains a little over 70% of our variation in attendances. I could add a few things in future if anyone has any ideas. I hope this has been of interest.
0 commentaires:
Enregistrer un commentaire