Certified Legendary Thread Race for the flag, in squiggly lines

[Final Siren]

In the "also rans" chart which team is to the far right above West Coast 94? Is it Saints 2009?

That is Collingwood 2011, who were super-defensive under Mick Malthouse. The team that beat them is Geelong 2011 at the very top of the chart.

 

[The Swert]

Interesting...

Why is Essendon's offensive rating below Sydney and North when they have outscored both teams this year?

 

[Final Siren]

Are the two crosses on the far right Adelaide in 2005 & 2006?

One is, yes. I've labelled a few notable misses:

 

[SBD Gonzalez]

So, what we can establish from this mess is as follows;

Teams that score more than their opposition are more likely to win.

Teams that restrict their opposition to lower scores are less likely to lose.

Cool.

What's amazing is how few seem to grasp this.

 

[Rushed_Behind]

Mr Squiggle on acid????

 

[Final Siren]

Interesting...

Why is Essendon's offensive rating below Sydney and North when they have outscored both teams this year?

It's because while Essendon has won games, it hasn't kicked many big scores against good defenses, especially recently. Looking at Round 18, for example, Essendon kicked an average score (87) against an average defense (Hawks), while Sydney scored heavily (110) against a slightly better defense (Richmond), and North Melbourne scored very heavily (150) against a poor defense.

Essendon has been higher offensively than North & Sydney during the season, even just a few weeks ago, but right now is lower.

There are always going to be situations where a lower-ranked team beats a higher-ranked one, of course, because that's footy. The model won't say that automatically means the lower-ranked one must be better; it's just another data point.

 

[djboonie]

Great stuff FS. Really really interesting.

I was just looking at defensive ratings this morning and the 2 stand outs are Sydney and Freo. Interestingly Syd are 9th on the list for conceding Inside 50's but 1st/2nd in Conceding Points Per Shot, Scoring shots conceded per Inside 50, and Average Scoring shots conceded.

Team    avgS/s    avgI50    PtsAg    PPS    S/sI50
Freo    19.94    44.53    1184    3.49    0.45
Coll    22.65    46.76    1460    3.79    0.48
Haw    22.59    47.47    1409    3.67    0.48
Port    22.71    47.53    1436    3.72    0.48
Geel    21.82    47.71    1376    3.71    0.46
Rich    23.29    48.76    1401    3.54    0.48
Kang    23.76    48.82    1464    3.62    0.49
Ess    23.65    48.88    1452    3.61    0.48
Syd    21.00    48.94    1232    3.45    0.43
Carl    25.47    49.41    1478    3.41    0.52
WCE    24.53    50.29    1622    3.89    0.49
Adel    23.88    52.00    1506    3.71    0.46
WB    28.12    52.65    1738    3.64    0.53
STK    27.12    52.94    1681    3.65    0.51
Bris    27.47    54.35    1702    3.64    0.51
GCS    26.29    55.12    1592    3.56    0.48
Melb    34.24    59.12    2117    3.64    0.58
GWs    35.24    60.82    2314    3.86    0.58

 

[tales129]

I'd like to know the initial model parameters in regards to finding a teams offensive and defensive percentage,

Is it just a percentage of the total points for each match,

Obviously once you've got the initial data points then you can use them to predict gsmes

 

[Lord Eddard]

I dont get the x & y indexes in the first graph? What are the percentages of?

 

[The Passenger]

My team is looking fairly good then. I approve.

i can't even see the west coast flag.

but i still vote that this should be the new AFL table as displayed in the Herald Sun every Monday. None of this 4 points for a win bollox. We demand charts!!!

 

[The Passenger]

I'd like to know the initial model parameters in regards to finding a teams offensive and defensive percentage,

Is it just a percentage of the total points for each match,

Obviously once you've got the initial data points then you can use them to predict gsmes

second this.

in round 1 every either starts equal or from the end of the season prior, but that is not entirely accurate due to so many different squads.

also when you say Hawthorn has a defence rating of 74%, that is 74% of what?

i can see where it's going, but getting the actual particulars is a bit harder to grasp.

 

[NUFC Tiger]

Good stuff Final Siren - fan of your stat work

 

[lamaros]

It's because while Essendon has won games, it hasn't kicked many big scores against good defenses, especially recently. Looking at Round 18, for example, Essendon kicked an average score (87) against an average defense (Hawks), while Sydney scored heavily (110) against a slightly better defense (Richmond), and North Melbourne scored very heavily (150) against a poor defense.

Essendon has been higher offensively than North & Sydney during the season, even just a few weeks ago, but right now is lower.

There are always going to be situations where a lower-ranked team beats a higher-ranked one, of course, because that's footy. The model won't say that automatically means the lower-ranked one must be better; it's just another data point.

So, it terms of Premiership chances it all down to form at the end of the season, yes?

What's the weighting?

 

[SBD Gonzalez]

Mr Squiggle on acid????

As Mr Squiggle used to say, you've gotta draw the line somewhere.

 

[Glasgow Hawk]

 

[Final Siren]

I'd like to know the initial model parameters in regards to finding a teams offensive and defensive percentage,

Is it just a percentage of the total points for each match,

Obviously once you've got the initial data points then you can use them to predict gsmes

I start the year before (so Round 1, 2012, for this chart) with all teams ranked 0.50 (or 50%) in both OFF and DEF, and go from there.

One season's worth of background has turned out to be a better predictor than 0 or 2 or 3 or 10, for whatever reason.

 

[BM123]

uhh wheres Carlton in the graph?

 

[Final Siren]

I dont get the x & y indexes in the first graph? What are the percentages of?

Nothing, really. I probably shouldn't have a % sign there, but it's kind of nice to think of it in terms of an ideal team.

In fact, the scores are entirely arbitrary and only have meaning in relation to each other. They don't measure any real-world absolute. If I'd assigned starting values of 100% or 100 to teams instead, they would sit in the same places on the chart relative to each other but their values would be different.

 

[Final Siren]

i can't even see the west coast flag.

It's hiding underneath Essendon.

uhh wheres Carlton in the graph?

Carlton is the caped crusader in the middle of the rainbow spaghetti!

The real challenge is spotting Collingwood, which is peeking over the top of Adelaide.

 

[stellation]

So, what we can establish from this mess is as follows;

Teams that score more than their opposition are more likely to win.

Teams that restrict their opposition to lower scores are less likely to lose.

Cool.

I think you're reading too much into this.

 

[Brain Power]

Obviously Hawthorn aren't as good defensively as Sydney or Freo, but I was very surprised to see them ranked the same as North, who are absolutely woeful at defending.

 

[Jones2ByrneJones]

A little while ago, I remember someone on the Port board, think it may have been RussellEbertHandball, said the best way to know which team would challenge for the premiership the next year is to look at % (and other factors ofc, but he stressed %, as it's a better indicator than W/L).

This sort of demonstrates that.

 

[Simon_Nesbit]

Superb stuff. Don't be discouraged by many, it's quite intuitive and simple to follow.

Would love to see the timelines for individual clubs over multiple years (eg comparing Hawks 08,10,12,13, Geelong 07-13, etc).

 

[BM123]

Surprised to see Essendon behind so many other teams in both categories and still be quite far up on the ladder.

 

[Final Siren]

So, it terms of Premiership chances it all down to form at the end of the season, yes?

Just looking over the charts, I don't see many examples of teams dashing into the ideal zone at the end of the season. Premiers tend to spend a lot of time in there.

The only recent counter-example I see is Hawthorn in 2008, which didn't really get in there until the start of the finals series. But according to this model, Hawthorn shouldn't have won at all. The Cats lived in that ideal zone all season.

What's the weighting?

As per the OP, it's a weighted average that works like this:

1. Most recent game: 9%
2. 2nd most recent game: 91% of 9% = 8.1%
3. 3rd most recent game: 91% of 8.1% = 7.4%
4. 4th most recent game: 91% of 7.4% = 6.7%
5. 5th most recent game: 91% of 6.7% = 6.1%
6. etc

And it's 91% and not some other value just because that has been the best predictor over the last 20 years.