14,583

[kaysee]

...
And that kind folks of the Bay is how a trend analysis for this kind of data is done (properly). D_One knows this but selectively chose misleading numbers.
...

All nicely done but still flawed if you're drawing comparisons across the 2 teams.

 

[Pilchard_Adams]

Oh that reminds me of yet another gem ....

"Retrograde Improvement"

Ha

RETROGRADE IMPROVEMENT ...is that what Port have been doing ..improving in a retrograde fashion.????

Two year Lags, Belting the living suitcase out the Gold Coast, No Improvement for the Krows, You had Daniel Bass and we didn't.........so much fail from one person.

Lag theory thread should be bumped

 

[subaru]

Lag theory thread should be bumped

Agree.....
Knock yourself out Pilch - I can't use this new search function

I can't even remember who was meant to be lagging who ????
Either way - it's another theory that turned to shit.

 

[Asgardian]

Regression analysis has always been used on time based data.

Time based regression analysis



Can you do a 5 year moving average? A linear regression also shows the obvious downward trend.



No they also explain the correct technique.

Mate, you had them all beat at least 25 pages back, all you're doing now is twisting the knife & they are too dumb to feel it.

 

[Geoffa32]

Bagged your misses last night... that was quality.

Was she better than Mrs Pearce?

 

[Mattrox]

All nicely done but still flawed if you're drawing comparisons across the 2 teams.

Never drew direct comparisons. Just looking at trends.

 

[D_One]

View attachment 2164 To do an analysis for Crowds you need to select the correct method.

I am looking to determine a change in the "short term" crowd trends. To do this a short term moving average is required. I am looking for a cross over of the short term average over the line of the actual data. The Crows season data crosses up over the two year moving average line. This shows that the short term downward trend has broken this season.

Look at the Port Power graph, no crossover signalling a change from the recent downward trend.

Thats a good graph. It highlights the Camry Crows decline real well. You can even see that from 97 onwards the declines are parallel.

However as the third order polynomial regression shows there is a nose dive after 2008. In far there are two distinct phases to the Camry Crows crowd numbers, pre 2003 and post 2003.

Can you do me a favour and plot upper and lower bounds to these two seperate phases.

There's a good minion.

 

[kaysee]

Never drew direct comparisons. Just looking at trends.

Even so... the trending is flawed do to the dependent being influenced by other independents that haven't been filtered out of the data set. That either Port or the Crows have not played the same games, on the same days, against the same opposition, or in the same economic climate mean that the trending is devalued as these all have significant relationships to the dependent value. Reducing the data sets to factor these values in results in far too small a sample size on which to draw accurate trending conclusions.

my .02 take or leave.

 

[D_One]

P.S.

D_One if those 2 graphs were presented for you as possible share investments which one would you jump on? Red line or Blue line.

Correct answer, you would jump on the blue line asap because if you waited the share price would recover back to historical levels and you would make a tidy capital gain.

Folks D_One will refuse to answer this question.

Correct answer: neither

BTW I've helped you out on plotting the two seperate phases.

 

[Mattrox]

So this is why polynomial regression is not suitable to analyse trends in football crowds.


Lets first try to use these lines of best fit to model football crowds.

First I used the, now familiar, Crows graph. I used a 3rd order polynomial like D_One suggests shows our downwards trend in crowds. I got Excel to place the formula of the line on the graph so it can be seen. I repeated the process for a 5th order polynomial. I did this because the 5th order fit the original line better.

The thing is, when you do a regression there is a value called R-squared. This tells you how accurate your regression line fits the original line that we got from real data. If R-squared = 1 then it is a perfect fit. R-squared values of 50 or below mean that the regression is a pretty useless fit for the data.

We see that the R-squared for the 3rd order polynomial is really low. That causes us to realise that this is not a good model for our data.

Next we see that the 5th order polynomial has a better R-squared value (closer to 1). The actual value means to line is not a very good fit. But since it is closer we might run with it.






I then used this line to calculate estimated crowds from 1991 to 2026. After all what use is a trend if we can't have a prediction about the next few years? I then graphed these predictions.

First, lets look at the crowd numbers if they continued to follow the so called "trends" described by the polynomials.


Holy crap!

If the 3rd order polynomial trend is continued the Crows will draw crowds of -41169! Phew, luck the R-squared value shows us this is not going to happen because the "trend line" doesn't fit our data at all.

Hey this is cool, the 5th order polynomial predicts that we will fill Adelaide Oval to capacity and overflowing by 2015. And in 2025 we will need to build a one million seat stadium to seat all the fans!! What does the graph look like?



Ok, reality check. Neither of these polynomials actually are trends. They just happen to be lines that kind of fit the original data. In addition football seasons are not a measure of time, it might happen to be season "2012" but really it is "the Crows 22nd Season". Time is measured in seconds, minutes, hours etc even years, but not in football seasons.

But if you are deluded you can measure the crowd trends with polynomials if you like but the 5th order is a better fit and is showing an upward trend for the Crows.

 

[Mattrox]

Even so... the trending is flawed do to the dependent being influenced by other independents that haven't been filtered out of the data set. That either Port or the Crows have not played the same games, on the same days, against the same opposition, or in the same economic climate mean that the trending is devalued as these all have significant relationships to the dependent value. Reducing the data sets to factor these values in results in far too small a sample size on which to draw accurate trending conclusions.

my .02 take or leave.

And that is what Crows fans have been trying to get across. Another example is we have home 2 games against expansion teams this year that will effect season averages. Someone refuses to take these different conditions into account.

And trends are not perfect, they are a guide. We all know that onfield success will instantly change the trends and often the largest rise is seen in the season following the first change in the trends. In addition, moving averages lag behind the real data so there are limitations.

Someone was also trying to bait Crows fans that the numbers were "Nose-diving" by "reducing the data sets[removing Showdowns] (in a poor attempt) to factor these values in results in far too small a sample size on which to draw accurate trending conclusions." Just that taking out showdowns are the only games that have the same factors in common.

 

[Mattrox]

Correct answer: neither

BTW I've helped you out on plotting the two seperate phases.

All well and good but you missed the change in short term trend as signalled by the "crossover" event.

Again selective y axis etc etc etc.

BTW your answer is a cop out.

But at least now we know that you would not advocate new football fans to get on Port Power, their stocks are dropping and have no future. We heard it first from you here!!

 

[Mattrox]

Correct answer: neither

BTW I've helped you out on plotting the two seperate phases.

And we know this is not a correct trend because all crowd figures from 1991 to 2009 are in a narrow range. Infact to start a downward trend before 2009 is a non-sense (where you choose to break the graph). You could put in sections in a number of places to paint what ever picture you want. You can present that graph, and I can present mine and we will see in time who build the best model of Crows crowds.

The greatest satisfaction will be next year when it will be shown that Crows crowds are not "nose-diving" as you assert.

 

[D_One]

So this is why polynomial regression is not suitable to analyse trends in football crowds.


Lets first try to use these lines of best fit to model football crowds.

First I used the, now familiar, Crows graph. I used a 3rd order polynomial like D_One suggests shows our downwards trend in crowds. I got Excel to place the formula of the line on the graph so it can be seen. I repeated the process for a 5th order polynomial. I did this because the 5th order fit the original line better.

The thing is, when you do a regression there is a value called R-squared. This tells you how accurate your regression line fits the original line that we got from real data. If R-squared = 1 then it is a perfect fit. R-squared values of 50 or below mean that the regression is a pretty useless fit for the data.

We see that the R-squared for the 3rd order polynomial is really low. That causes us to realise that this is not a good model for our data.

Next we see that the 5th order polynomial has a better R-squared value (closer to 1). The actual value means to line is not a very good fit. But since it is closer we might run with it.
View attachment 2179

And if you used a 20th order polynomial regression what would your R-squared be?

A one year upturn is not a trend. The form of a team adds noise to the underlying trend.

There is no doubt that the underlying trend for the Camry Crows is down. This years ever so slight upturn due being top four is only a bit of noise in the overall downward trend.



I then used this line to calculate estimated crowds from 1991 to 2026. After all what use is a trend if we can't have a prediction about the next few years? I then graphed these predictions.

First, lets look at the crowd numbers if they continued to follow the so called "trends" described by the polynomials.
View attachment 2180

Holy crap!

If the 3rd order polynomial trend is continued the Crows will draw crowds of -41169! Phew, luck the R-squared value shows us this is not going to happen because the "trend line" doesn't fit our data at all.

Hey this is cool, the 5th order polynomial predicts that we will fill Adelaide Oval to capacity and overflowing by 2015. And in 2025 we will need to build a one million seat stadium to seat all the fans!! What does the graph look like?

View attachment 2181

Ok, reality check. Neither of these polynomials actually are trends. They just happen to be lines that kind of fit the original data. In addition football seasons are not a measure of time, it might happen to be season "2012" but really it is "the Crows 22nd Season". Time is measured in seconds, minutes, hours etc even years, but not in football seasons.

But if you are deluded you can measure the crowd trends with polynomials if you like but the 5th order is a better fit and is showing an upward trend for the Crows.

 

[Sydney Bloods]

Ok now I'm starting to think D_one works for the government.

Spin baby spin.

Tell us more about how low there crowds really are, simple stat's don't lie unlike your fancy circle jerk work.

28,261 is there lowest home crowd that's level with Carlton for the same game.

The only interstate team in The comp with higher crowds then the crows are the eagles.

Keep digging.

 

[kaltmeibel]

Thats a good graph. It highlights the Camry Crows decline real well. You can even see that from 97 onwards the declines are parallel.

However as the third order polynomial regression shows there is a nose dive after 2008. In far there are two distinct phases to the Camry Crows crowd numbers, pre 2003 and post 2003.

Can you do me a favour and plot upper and lower bounds to these two seperate phases.

There's a good minion.

Yet we still get profit making crowds. Oh thats right we have a dynamic CEO.

 

[Mattrox]

Ok now I'm starting to think D_one works for the government.

Spin baby spin.

Tell us more about how low there crowds really are, simple stat's don't lie unlike your fancy circle jerk work.

28,261 is there lowest home crowd that's level with Carlton for the same game.

The only interstate team in The comp with higher crowds then the crows are the eagles.

Keep digging.

The graphing is all well and good to detect trends. All D's split graph shows is that there was a drop in crowd numbers no shit sherlock. The trickf you are serious about trends is to detect a change in trend. In share analysis it is best to detect this early. I think that I have shown early signs of a change in trends,


But all this technical analysis needs to be balanced with the underlying fundamental. What factors are influencing home crowds. I believe that playing expansion teams at home draws lower crowds. That can be investigated,


D clearly knows how to use these techniques well, well enough to be very selective with the info he has presented. His split graph is not wrong per se, it just isn't the whole story. But look how long it took before he produced something that was not completely misleading. And look how long he persisted with arguing that polynomial curves are appropriate when they are not. Even straight line trends are not appropriate really, but they are simple and it is tricky to visually show that they are wrong.

 

[Mattrox]

Even if you do decide to use polynomials the 5th order is a better fit, and shows the upward trend. If you subscribe to polynomial methods , which I don't for AFL crowds, the 'upward trend' has started.

 

[Goat 53]

Strange business strategy that Poort Power have:

Deliberately run your franchise into the ground to cripple your stakeholders.

 

[D__One]

GWS could only manage 6,811 at Skoda Stadium today. Thats less than half of 14583

 

[Elite Crow]

GWS could only manage 6,811 at Skoda Stadium today. Thats less than half of 14583

But they managed to win

 

[subaru]

Can we please have a graph, " Crows /Power - Wins V Expansion Teams" tah, thx would be good if we could......

 

[Geoffa32]

GWS could only manage 6,811 at Skoda Stadium today. Thats less than half of 14583

So what did KT say when you asked him those tough questions?

 

[Mattrox]

With our fourth 40K+ Crowd at AAMI I think we can now say the downward trend has changed. The season average is now 37181.

If we look at the average without GWS to check numbers with previous years it is 38296. So on previos years, we are within cooee of the expected range.

D_One is wrong again, but I suspect he has more on his mind than the "nose-diving" Crows crowds.

It's up to you Port fans. Turn up or see your team fold.

Will you even get 10K to turn up to your next game?

 

[Geoffa32]

Is todays crowd one of the largest for this timeslot by any club this year?