Analysis Team Rankings 2017

[grumbleguts]

OK I've begun my team ranking analysis for this year, at two rounds there is barely enough information to point a stick at the competition so this will change drastically over the next few rounds and then settle down by about round 7. This system has correctly predicted the premier by round 12 in each season in the last 5 seasons except last season where Bulldogs were ranked third even though they finished 7th.

Hawthorn were ranked 6th last season even though they were top of the table for much of the second half of the season.

OK Here is 2017 after two rounds

Adelaide Crows 10.65345
Port Adelaide 10.34298
GWS Giants 9.6974
West Coast Eagles 8.019154
Richmond 7.858016
Melbourne 7.684356
Essendon 7.54201
Western Bulldogs 7.450269
Geelong Cats 6.838813
Hawthorn 6.330111
Sydney Swans 5.855137
Collingwood 5.723959
North Melbourne 5.655256
St Kilda 5.597797
Brisbane Lions 5.228693
Carlton 4.990054
Gold Coast Suns 4.313363
Fremantle 3.733836


Rankings After Round 3
GWS Giants 14.9
Adelaide Crows 13.839
Port Adelaide 12.893
Richmond 10.788
Geelong Cats 10.372
Gold Coast Suns 9.943
West Coast Eagles 9.804
Melbourne 9.75
St Kilda 8.473
Essendon 8.282
Western Bulldogs 8.274
Carlton 8.128
North Melbourne 7.67
Collingwood 7.566
Sydney Swans 7.393
Brisbane Lions 7.075
Fremantle 6.689
Hawthorn 6.249

Rankings after round 4
GWS Giants 19.737
Adelaide Crows 17.899
Geelong Cats 16.726
Port Adelaide 16.101
Richmond 15.15
West Coast Eagles 13.675
Gold Coast Suns 13.074
Melbourne 12.631
St Kilda 11.627
Western Bulldogs 11.061
North Melbourne 10.804
Carlton 10.25
Essendon 10.109
Fremantle 9.931
Collingwood 9.81
Sydney Swans 9.451
Brisbane Lions 8.856
Hawthorn 7.37

Team Rankings after round 5
GWS Giants 24.525
Adelaide Crows 24.378
Port Adelaide 23.713
Geelong Cats 21.741
Richmond 17.915
Gold Coast Suns 16.056
Melbourne 15.261
West Coast Eagles 15.039
Western Bulldogs 14.332
St Kilda 13.862
Essendon 13.692
North Melbourne 13.263
Fremantle 13.111
Hawthorn 12.521
Sydney Swans 12.09
Collingwood 12.007
Carlton 11.621
Brisbane Lions 10.881

Team Rankings after round 6
Adelaide Crows 31.07
Port Adelaide 29.333
GWS Giants 27.876
Geelong Cats 24.354
Richmond 20.709
St Kilda 20.221
West Coast Eagles 19.808
Melbourne 19.634
Western Bulldogs 18.498
Gold Coast Suns 18.448
Collingwood 17.147
North Melbourne 16.545
Essendon 16.136
Fremantle 15.32
Carlton 14.791
Sydney Swans 14.714
Hawthorn 14.252
Brisbane Lions 12.867

Round 7 rankings

Port Adelaide 33.477
Adelaide Crows 33.457
GWS Giants 32.045
Geelong Cats 28.137
West Coast Eagles 25.8
St Kilda 25.606
Richmond 24.042
North Melbourne 24.041
Gold Coast Suns 23.631
Melbourne 22.901
Western Bulldogs 22.645
Fremantle 19.976
Collingwood 19.627
Carlton 19.206
Sydney Swans 19.089
Essendon 18.199
Hawthorn 17.866
Brisbane Lions 14.987


Rankings after Round 8
Port Adelaide 43.288
Adelaide Crows 36.582
GWS Giants 35.876
Geelong Cats 31.734
West Coast Eagles 30.51
St Kilda 30.188
Melbourne 30.149
Richmond 27.72
Western Bulldogs 26.204
North Melbourne 25.985
Gold Coast Suns 25.656
Sydney Swans 24.46
Fremantle 24.157
Collingwood 24.025
Essendon 23.111
Carlton 22.951
Hawthorn 22.125
Brisbane Lions 17.404

Round 9
Port Adelaide 43.272
Adelaide Crows 37.287
GWS Giants 34.559
Geelong Cats 32.109
Melbourne 29.852
West Coast Eagles 28.806
St Kilda 28.382
Richmond 28.003
North Melbourne 26.83
Sydney Swans 26.699
Essendon 26.607
Western Bulldogs 26.218
Fremantle 26.097
Gold Coast Suns 25.697
Collingwood 24.8
Carlton 22.447
Hawthorn 21.812
Brisbane Lions 16.971

Round 10
Adelaide Crows 44.993
Port Adelaide 44.936
GWS Giants 37.007
Geelong Cats 35.087
Melbourne 32.907
West Coast Eagles 30.929
Richmond 30.867
Western Bulldogs 30.558
North Melbourne 29.533
St Kilda 29.419
Essendon 28.505
Sydney Swans 28.282
Gold Coast Suns 27.46
Collingwood 27.449
Fremantle 26.787
Hawthorn 24.422
Carlton 24.119
Brisbane Lions 18.425

Round 11
Port Adelaide 49.547
Adelaide Crows 47.649
GWS Giants 40.025
Geelong Cats 39.438
Melbourne 35.957
Richmond 34.504
Western Bulldogs 33.234
West Coast Eagles 32.982
St Kilda 31.886
Essendon 31.234
North Melbourne 31.005
Sydney Swans 30.529
Collingwood 30.282
Gold Coast Suns 30.133
Fremantle 28.641
Carlton 26.317
Hawthorn 25.735
Brisbane Lions 20.121

Round 12

Adelaide Crows 48.726
Port Adelaide 46.249
Geelong Cats 39.327
GWS Giants 38.143
Melbourne 35.815
Essendon 35.543
Richmond 35.088
West Coast Eagles 32.297
Western Bulldogs 32.125
Sydney Swans 31.294
St Kilda 30.95
North Melbourne 30.709
Collingwood 30.401
Gold Coast Suns 27.782
Fremantle 27.493
Carlton 27.364
Hawthorn 26.501
Brisbane Lions 23.044


Round 13
Adelaide Crows 48.726
Port Adelaide 45.176
Geelong Cats 38.383
GWS Giants 38.359
Melbourne 37.589
Essendon 35.577
Richmond 34.601
West Coast Eagles 33.264
Sydney Swans 32.044
St Kilda 31.573
Western Bulldogs 31.315
Collingwood 30.799
North Melbourne 30.509
Carlton 27.964
Fremantle 27.828
Gold Coast Suns 27.488
Hawthorn 26.927
Brisbane Lions 23.325

Round 14
Adelaide Crows 52.608
Port Adelaide 49.3
GWS Giants 42.283
Geelong Cats 41.042
Melbourne 40.279
Essendon 38.569
Richmond 37.89
West Coast Eagles 36.341
Sydney Swans 35.345
St Kilda 34.909
Western Bulldogs 34.397
Collingwood 33.237
North Melbourne 33.05
Hawthorn 31.183
Carlton 31.104
Fremantle 30.77
Gold Coast Suns 28.573
Brisbane Lions 25.3

Round 15

Adelaide Crows 51.496
Port Adelaide 48.101
GWS Giants 41.911
Geelong Cats 40.707
Melbourne 38.732
Richmond 38.666
Essendon 38.152
Sydney Swans 37.357
West Coast Eagles 36.584
St Kilda 34.71
Western Bulldogs 34.67
North Melbourne 32.954
Collingwood 32.651
Hawthorn 32.036
Carlton 31.556
Fremantle 30.656
Gold Coast Suns 29.914
Brisbane Lions 26.03


Round 16

Adelaide Crows 58.107
Port Adelaide 53.964
Geelong Cats 46.215
GWS Giants 45.414
Sydney Swans 43.042
Melbourne 42.567
Essendon 42.382
St Kilda 40.802
Richmond 40.625
West Coast Eagles 39.596
Western Bulldogs 37.104
North Melbourne 36.748
Hawthorn 36.094
Collingwood 35.982
Carlton 35.255
Fremantle 33.568
Gold Coast Suns 31.955
Brisbane Lions 28.112

Round 17
Adelaide Crows 59.624
Port Adelaide 55.106
Essendon 45.082
Sydney Swans 44.097
GWS Giants 44.05
Geelong Cats 42.456
Richmond 40.179
Melbourne 39.924
West Coast Eagles 39.199
St Kilda 39.022
Hawthorn 37.382
Western Bulldogs 36.898
Collingwood 35.52
North Melbourne 34.979
Carlton 34.574
Fremantle 33.608
Gold Coast Suns 31.283
Brisbane Lions 28.498

 

[OnTheRocks]

Swines not playing finals! AFL won't let that happen.

 

[Gralin]

As grumble said it's round two

 

[Dave Id]

Crows and Port playing the Grand Final at the MCG.
Don't know what to say.

 

[grumbleguts]

It's amazing how thhe maths works out, GWS were bottom of the table last week, and after 1 round, the analysis gives a lot of "undefined" errors in the rankings. It works like this:

If A beats B and B beats C, then it will work out that A will beat C.
BUT
If A beats B and C beats D there is not enough information to say who will win out of anyone who hasn't played. D might be the second best team in the comp.

Two rounds is just enough to get an inkling. But very hazy. After 7 rounds things will settle. It uses VERY different maths to the squiggle, but in most cases aligns nicely.

It also can predict individual teams beating those ranked higher than them. But that needs much longer posts.

 

[grumbleguts]

Edited the forst port to include the ranking after round 3. Not looking the best at the moment.

 

[Gralin]

Edited the forst port to include the ranking after round 3. Not looking the best at the moment.

It is what it is.

 

[grumbleguts]

It is what it is.

So, it seems, is my typing.

(in the dark can't see the keyboard.)

 

[grumbleguts]

OP updated to include round 4.

Not pretty reading.

 

[grumbleguts]

OP edited to include round 5 data.

 

[Crankyhawk]

Is the method proprietary or just very complex to explain?

 

[grumbleguts]

It's getting interesting nw, there's enough data for little ripples to affect distant teams. For example if we beat St Kilda next week by doubling their score, say 112 to 56, and eveeryone else plays exactly according to their current ranking, we end up ranked 8th and Adelaide over-take GWS purely based on the result of the Hawthorn game, because adelaide beat hawthorn in round 2.

 

[OzJohnnie]

Did you post your methodology last year, grumbleguts? I need a refresher - so I can figureout how Hawthorn gets to the top.

 

[Mitchell54]

Now we need to smash St Kilda and West Coast needs to smash everyone so it makes us look good.
Is that how it works?

 

[grumbleguts]

Now we need to smash St Kilda and West Coast needs to smash everyone so it makes us look good.
Is that how it works?

Yes, essentially.

 

[grumbleguts]

Did you post your methodology last year, grumbleguts? I need a refresher - so I can figureout how Hawthorn gets to the top.

This explains it, but it is far more complicated because it is 18 teams and the score is taken into account:

 

[grumbleguts]

Did you post your methodology last year, grumbleguts? I need a refresher - so I can figureout how Hawthorn gets to the top.

If you've watched the video, this is my secoond order dominance matrix after 5 rounds this season.



but I also use the 3rd and 4th order matrices as well. (you can see the top of the third order matrix at the bottom.) I don't start with ones and zeros though I use the score ratio for the first order matrix, so the margin is taken into account.

 

[Mitchell54]

Yes, essentially.

Suppose it's good then that they play in Perth this week

 

[grumbleguts]

Suppose it's good then that they play in Perth this week

Although they are playing fremantle who don't have a high ranking themselves, so not a lot of gain there.

 

[Mitchell54]

Although they are playing fremantle who don't have a high ranking themselves, so not a lot of gain there.

Not much gain for us but they wouldn't do too well if they lost to Fremantle. They're our only bargaining chip

 

[OzJohnnie]

This explains it, but it is far more complicated because it is 18 teams and the score is taken into account:

Ok, thanks. So we need blowout wins against highly ranked teams to really move up. And we need those teams to keep winning. But with Adelaide and Geelong already beating us, every win we get pushes them higher. Hmm...

A couple questions, if you don't mind:

* Early season wins carry the same weight as later wins? Intuitively, I would expect this to overweight early good form (North last year) and underweight late season good form (hopefully us this year). But except for last year this method has hit the winner after round 7 every time? How many years have you tested that on? I'm curious because we're shot at round 7 with this and I want hope. Show me you can be wrong!

* Do you use straight scoring differential with no adjustments? Again, intuitively, I would think this would lead to overweighting blowouts (which almost always blow further when one team drops it's head - us against Geelong and GC this year) while underweighting the close clash of giants (or at least evenly weighted teams) which can really set the tone for dominance. Of course, the model has been right after round 7, but do you have thoughts on this?

* One last question: third and fourth vectors? Particularly the fourth. Does it make a difference to the rankings when the games are three teams removed?

Thanks, grumbleguts. I appreciate your efforts here.

 

[grumbleguts]

Ok, thanks. So we need blowout wins against highly ranked teams to really move up. And we need those teams to keep winning. But with Adelaide and Geelong already beating us, every win we get pushes them higher. Hmm...

A couple questions, if you don't mind:

* Early season wins carry the same weight as later wins? Intuitively, I would expect this to overweight early good form (North last year) and underweight late season good form (hopefully us this year). But except for last year this method has hit the winner after round 7 every time? How many years have you tested that on? I'm curious because we're shot at round 7 with this and I want hope. Show me you can be wrong!

* Do you use straight scoring differential with no adjustments? Again, intuitively, I would think this would lead to overweighting blowouts (which almost always blow further when one team drops it's head - us against Geelong and GC this year) while underweighting the close clash of giants (or at least evenly weighted teams) which can really set the tone for dominance. Of course, the model has been right after round 7, but do you have thoughts on this?

* One last question: third and fourth vectors? Particularly the fourth. Does it make a difference to the rankings when the games are three teams removed?

Thanks, grumbleguts. I appreciate your efforts here.

I weight the third and fourth order vectors very lowly. Games are weighted equally throughout the year. The weightings I use come from an assignment I gave a senior maths class, where i gave them the first 7 rounds of two consecutive afl seasons, but I used meerkats, gorillas, flamingos etc instead of tigers, hawks, kangaroos etc. SO they couldn't look up the actual seasons. Then I got them to predict the final ladder positions of each of the teams, using different weightings for each vector order. The combination that came closest to the final ladder positions in both seasons is what I use, which I check against the final ladder each season and refine slightly.

 

[OzJohnnie]

I weight the third and fourth order vectors very lowly. Games are weighted equally throughout the year. The weightings I use come from an assignment I gave a senior maths class, where i gave them the first 7 rounds of two consecutive afl seasons, but I used meerkats, gorillas, flamingos etc instead of tigers, hawks, kangaroos etc. SO they couldn't look up the actual seasons. Then I got them to predict the final ladder positions of each of the teams, using different weightings for each vector order. The combination that came closest to the final ladder positions in both seasons is what I use, which I check against the final ladder each season and refine slightly.

 

[grumbleguts]

Ok, thanks. So we need blowout wins against highly ranked teams to really move up. And we need those teams to keep winning. But with Adelaide and Geelong already beating us, every win we get pushes them higher. Hmm...

A couple questions, if you don't mind:

* Early season wins carry the same weight as later wins? Intuitively, I would expect this to overweight early good form (North last year) and underweight late season good form (hopefully us this year). But except for last year this method has hit the winner after round 7 every time? How many years have you tested that on? I'm curious because we're shot at round 7 with this and I want hope. Show me you can be wrong!

* Do you use straight scoring differential with no adjustments? Again, intuitively, I would think this would lead to overweighting blowouts (which almost always blow further when one team drops it's head - us against Geelong and GC this year) while underweighting the close clash of giants (or at least evenly weighted teams) which can really set the tone for dominance. Of course, the model has been right after round 7, but do you have thoughts on this?

* One last question: third and fourth vectors? Particularly the fourth. Does it make a difference to the rankings when the games are three teams removed?

Thanks, grumbleguts. I appreciate your efforts here.

Sorry, I realised I didn't answer your last question, the multiplying factor is very small here, because it should have a huge effect, when there is little data. It makes a difference to the rankings once there are enough games for there to be three games distance between two teams on the grid. A beats B B beats C C beats D it will tell us that A beats D, (but it is so far away from the actual direct empirical result, it isn't weighted much.

But you need a few rounds for it to kick in, by way of example, (similar to the video imagine three teams) A beats B and B beats C.

The first order matrix (directly from the data looks like this

......A..B..C
A....0..1...0
B....0..0...1
C....0..0...0

If that matrix is squared you get the second order matrix

......A..B..C
A....0..0...1
B....0..0...0
C....0..0...0

A beats C

 

[grumbleguts]

There's an article on the main bigfooty page stating that Port Adelaide enters the top four on the squiggle. My system has had them in the top four from the beginning.