2013 AFL Power Rankings

[Richo83]

As much as I like the surge in Richmond's rankings, some of them seem iffy. Chiz your rakings say that Richmond are 3rd, even though they've lost to your 2nd, 4th, 6th and 7th ranked team and have only beaten 8th and below. The issue of whether Richmond should be a top four ranked side does seem to be in part determined by their ability to beat top four sides. It is possible to rank such a team 6-8th, but 3rd?

 

[Chiz]

As much as I like the surge in Richmond's rankings, some of them seem iffy. Chiz your rakings say that Richmond are 3rd, even though they've lost to your 2nd, 4th, 6th and 7th ranked team and have only beaten 8th and below. The issue of whether Richmond should be a top four ranked side does seem to be in part determined by their ability to beat top four sides. It is possible to rank such a team 6-8th, but 3rd?

Your point is entirely valid. If I was to rank them subjectively, I'd have them below Geelong and Sydney and about equal with Essendon. However, I will make two points in regards to the rankings:

1. The rankings are self-correcting, which means that a team who is ranked too high will gain fewer points for wins and lose more points for losses - ultimately, the ratings will converge. Hawthorn are probably rated too highly at the moment because they haven't lost a game for a while and hence haven't lost any points. When they finally lose a game, they will probably drop 6 points or so.

2. This kind of relates to the first point, but the ratings are largely form-based. Richmond, for instance, had a big jump in points after defeating West Coast in Perth (and have remained steady with two wins since). Sydney and Geelong both lost a lot of ground after unexpected losses on the weekend.
If I was to take the average rating of each team over the last five rounds (for example), this would probably give a more accurate reflection of where each team stands, and would most likely see Richmond back around the 6-8th mark.

 

[Simon_Nesbit]

If we're sharing our own ladders, the best ladder is the "last home and away" against every team. For the established clubs, games against GC and GWS have been ignored. (thanks Darwi for the stats resource)

Ld Team .W .L D ..P Pct% 
01 Haw: 27 .3 0 108 140.8%
02 Swa: 21 .7 2 .88 134.9%
03 Cat: 22 .8 0 .88 119.9%
04 Mag: 21 .9 0 .84 110.8%
05 Doc: 20 .9 1 .82 111.7%
06 Eag: 18 12 0 .72 113.4%
07 Bom: 18 12 0 .72 110.3%
08 Cro: 14 16 0 .56 93.9%
09 Tig: 13 16 1 .54 103.5%
10 Kan: 13 17 0 .52 104.9%
11 Blu: 13 17 0 .52 100.1%
12 Sai: 11 19 0 .44 97.3%
13 Bul: 10 20 0 .40 79.3%
14 Lio: .8 22 0 .32 79.0%
15 Pow: .7 22 1 .30 77.0%
16 Sun: .7 22 0 .28 67.7%
17 Dem: .1 28 1 ..6 58.8%
18 Gia: .0 24 0 ..0 45.9%

Look how dominant Hawthorn have been - yet people still ridiculed Roby.

FWIW, our losses are home and away games against Geelong, and a game against Richmond - win those and we get the complete set - I don't think that has ever been done in the history of the game.

 

[mHmmm]

If we're sharing our own ladders, the best ladder is the "last home and away" against every team. For the established clubs, games against GC and GWS have been ignored. (thanks Darwi for the stats resource)

Ld Team .W .L D ..P Pct%
01 Haw: 27 .3 0 108 140.8%
02 Swa: 21 .7 2 .88 134.9%
03 Cat: 22 .8 0 .88 119.9%
04 Mag: 21 .9 0 .84 110.8%
05 Doc: 20 .9 1 .82 111.7%
06 Eag: 18 12 0 .72 113.4%
07 Bom: 18 12 0 .72 110.3%
08 Cro: 14 16 0 .56 93.9%
09 Tig: 13 16 1 .54 103.5%
10 Kan: 13 17 0 .52 104.9%
11 Blu: 13 17 0 .52 100.1%
12 Sai: 11 19 0 .44 97.3%
13 Bul: 10 20 0 .40 79.3%
14 Lio: .8 22 0 .32 79.0%
15 Pow: .7 22 1 .30 77.0%
16 Sun: .7 22 0 .28 67.7%
17 Dem: .1 28 1 ..6 58.8%
18 Gia: .0 24 0 ..0 45.9%

Look how dominant Hawthorn have been - yet people still ridiculed Roby.

FWIW, our losses are home and away games against Geelong, and a game against Richmond - win those and we get the complete set - I don't think that has ever been done in the history of the game.

It also demonstrates just how terrible Melbourne are. They almost have a complete set, just not the set you want.

 

[The_Ru]

If we're sharing our own ladders, the best ladder is the "last home and away" against every team. For the established clubs, games against GC and GWS have been ignored. (thanks Darwi for the stats resource)



Look how dominant Hawthorn have been - yet people still ridiculed Roby.

FWIW, our losses are home and away games against Geelong, and a game against Richmond - win those and we get the complete set - I don't think that has ever been done in the history of the game.

After Rd 13 2011 Geelong had the full set minus only Adelaide away. In Rd 14 they beat Adelaide by 52 points... at home.

To collect the full set is almost impossible because you need to maintain significant winning streaks AND have the fixture fall the right way for you. (You may not play a team at home for 2 seasons).

For example, for Hawthorn to get the full set they'll need to beat Geelong at home and away and Richmond without dropping a match against anyone else home or away in the meantime. Their next home match against Geelong may not occur until late 2014. I'd be willing to bet money they can't do it.

 

[SuperSub17]

After Rd 13 2011 Geelong had the full set minus only Adelaide away. In Rd 14 they beat Adelaide by 52 points... at home.

Just by memory, what about Collingwood away? Last away game to the Pies was the 2010 prelim when they got smashed. They were home when they won in round 8 2011.

 

[The_Ru]

Just by memory, what about Collingwood away? Last away game to the Pies was the 2010 prelim when they got smashed. They were home when they won in round 8 2011.

I'm not counting finals as they don't traditionally contribute to a ladder. Not sure if Simon_Nesbit was counting finals.

 

[Richo83]

If we're sharing our own ladders, the best ladder is the "last home and away" against every team. For the established clubs, games against GC and GWS have been ignored. (thanks Darwi for the stats resource)

I'm not sure that is the best ladder. If I am right, then you take the results from when the last time a team plays any other team home, and the last time a team plays a team away. So the last time Hawthorn played Brisbane at home was in 2012 round 12, a victory, whereas the last time Hawthorn played Brisbane away was in 2008 round 5, also a victory. So you take those two results and give Hawthorn two wins. This is a flawed system, any power rankings system which uses data from 5 years ago will be misleading or false.

If one is to use this method, then it seems more fair that you analyse the last two results against each side. At the least, this keeps the focus to more recent results.

 

[Simon_Nesbit]

I'm not sure that is the best ladder. If I am right, then you take the results from when the last time a team plays any other team home, and the last time a team plays a team away. So the last time Hawthorn played Brisbane at home was in 2012 round 12, a victory, whereas the last time Hawthorn played Brisbane away was in 2008 round 5, also a victory. So you take those two results and give Hawthorn two wins. This is a flawed system, any power rankings system which uses data from 5 years ago will be misleading or false.

If one is to use this method, then it seems more fair that you analyse the last two results against each side. At the least, this keeps the focus to more recent results.

That's not an error in the system, that's an error in the scheduling.

At least we don't rely on the scheduling to decide anything important....like finals or draft picks....oh, yeah.

 

[Richo83]

That's not an error in the system, that's an error in the scheduling.

You're relying on the scheduling to provide an accurate account of club performances. Given scheduling is uneven and unfair, it seems unlikely that your system will be too. Again, I don't see why you would only analyse the last home and away results against each team, it seems arbitrary and seems like an attempt to factor for the impact of playing away, which, given there's 10 Victorian teams and a number of games being sold interstate, seems like a confused goal.

At least we don't rely on the scheduling to decide anything important....like finals or draft picks....oh, yeah.

You use data from 2008 to base your ladder, the 2013 finals and draft wont be determined by using results from 2008. At least the draft and finals has historical parameters on what is relevant and what isn't.

 

[Stewart2Austin Reborn]

I guess this explains why Port are so low, because our wins have come largely due to luck and umpire assistance.

While I don't necessarily agree, at least your rankings make more sense now.

Also Roby, if you could use this season's fixture and whatever, what's the probability Port will make finals?

 

[Stewart2Austin Reborn]

Hey Chiz love your work. According to your rankings, what's the probability Port will make finals?

 

[Chiz]

Hey Chiz love your work. According to your rankings, what's the probability Port will make finals?

Cheers. I'd need to run a large number of simulations and record the results, but I did run a simulation where Port made the top 4, so there's always a chance! Closer to the end of the season I plan to record some simulation results and calculate all the odds for each team (premiership, top 4, finals).

 

[Simon_Nesbit]

You're relying on the scheduling to provide an accurate account of club performances. Given scheduling is uneven and unfair, it seems unlikely that your system will be too. Again, I don't see why you would only analyse the last home and away results against each team, it seems arbitrary and seems like an attempt to factor for the impact of playing away, which, given there's 10 Victorian teams and a number of games being sold interstate, seems like a confused goal.

I see it from the opposite side. The scheduling issue is one of the biggest influences on ladder position year to year, with flow on effects on clubs overall, marketing, tv exposure and players wanting to come to clubs. By looking at the last home and away, it removes the vagaries of the fixture.

You use data from 2008 to base your ladder, the 2013 finals and draft wont be determined by using results from 2008. At least the draft and finals has historical parameters on what is relevant and what isn't.

Ah, makes sense now. I'm not trying to predict a winner for 2013 (it's Hawthorn btw, ), it's a reflection of what the cumulative returns are from club to club. It doesn't factor in the bias of the fixture, but to do so you would have to introduce a Strength of Schedule factor, both passed and still to come.

In that regards, this ladder is more accurate than the one currently used. The AFL would be better served adopting this format, as the current fixture is highly manipulated for non-equitable purposes.

 

[Roby]

Haw and Freo have been top two in your power rankings since rd 10, ie the teams you think are the best premiership chances.

What has changed since rd 10 that has seen you pounce with a GF quinella bet now?

Why not at Rd10 when the value was much greater and your rankings first selected Freo and Haw as top two?

You're right it should've been done earlier.

I've been working on my own rankings system since the start of the season, and spent a fair bit of time making adjustments throughout the season to get it right. I thought I would put it in this thread, as it's not controversial enough to get attention in its own thread.

Unlike Roby's rankings, these are purely statistical - there are no subjective judgements made on each game (e.g. umpiring, injuries, betting markets, etc.). I gave each team a rating of 100 points at the start of 2010 (when Gold Coast and GWS entered the league they started on 50 points), and ratings are adjusted using a formula after each game (this takes into account probability of winning, the margin, home ground advantage, whether it is a high or low scoring game, etc.).

Using one of this week's games as an example, the ratings predict that North Melbourne have roughly a 90% probability of beating GWS, with a predicted margin of 60 points. If NM win by 60 points, their rating will go up roughly half a point, and GWS will go down half a point (i.e. not very much). If NM win by 100 points they gain 3.7 points (GWS lose 3.7). If they win by only 10 points NM actually lose 3.7 points. If GWS win by 10 points they will gain 8.6 points (NM lose 8.6).

In contrast, West Coast and Essendon are almost even money (probability of winning is 50%, which takes into account WCE have the home ground advantage, and a predicted margin of 0). If either team wins by 1 point they will gain 2 points to their rating (and the other team will lose 2 points). However, if one team wins by 50 points they gain 5 ratings points.

At the (just over) half way point of the season, the teams are rated as follows:

View attachment 21911

On a neutral venue, Hawthorn would have roughly a 65% chance of beating Fremantle, a 78% chance of beating North Melbourne, and an 83% chance of beating St Kilda, according to the above ratings (it is the difference in points that determines probability of winning).

As much as I disagree with the subjective nature of Roby's rankings, it is curious that Richmond are rated highly on both (although there is little separating RIC, GEE, SYD and ESS), although the main reason they are third is due to a sudden fall in both Geelong and Sydney's rating.
I also use these rankings to simulate the final ladder - HAW and GEE make the top four about 80% of the time, and the top four typically consists of three of HAW, GEE, FRE and SYD (occasionally ESS, RIC or COL make the top four, but it is highly unlikely that any other team will finish in the top four). Thus, despite Richmond's high rating, they have probably lost too many early games to give the top four a real crack.

I also calculate the betting odds for each game (this doesn't take into account the bookie's percentage). For round 14:

WCE: $2.00, ESS: $2.00
SYD: $1.37, CAR: $3.74
GC: $1.83, ADE: $2.20
PA: $2.28, COL: $1.78
MEL: $5.06, WB: $1.25
GEE: $1.55, FRE: $2.83
HAW: $1.20, BL: $6.05
NM: $1.13, GWS: $8.72
RIC: $1.25, SK: $5.02

Curious to hear your thoughts, Roby, and how much these figures differ to yours.

Realise this: when deciding how you weight matches or what kind of metric you will use, you are making a decision and therefore a subjective judgement on how to formulate the rankings.

I can't comment on your rankings until you rationalise in real world values why you have chosen your particular metric. Here is a link which rationalises how the power rankings are formulated. If you have any further questions regarding those I will probably answer them next week when I have more time.

As far as the betting lines are concerned, they are not taken into account. I just keep record of them and use them as comparison against the rankings.

In regards to the Tigers they are highly ranked because they still carry weight from last year. Tigers never finished lower than 12th in the rankings and as high as 6th. The wins against last year's grand finalist still are significant whether people like to admit this or not. In fact, Geelong are the only other team to have beaten both Sydney and Hawthorn in the last two seasons. And I probably ask you to rewatch the Fremantle v Richmond match back in round five. Forget that the Tigers blew the game at the end - monitor the umpiring decisions in that game and weight them. And maybe right now in alternate extremely similar universe to this current one, the Tigers would be in fourth place on the ladder.

 

[Richo83]

To be fair to Chiz Richmond are rated 3rd, which is a higher ranking than you currently have them, so I don't think either of you underrate Richmond.

Roby, according to your model, what have been the biggest upsets of the year? Is Port v Sydney the biggest?

 

[Roby]

To be fair to Chiz Richmond are rated 3rd, which is a higher ranking than you currently have them, so I don't think either of you underrate Richmond.

Roby, according to your model, what have been the biggest upsets of the year? Is Port v Sydney the biggest?

I understand what Chiz said, he said it was interesting, I was further elaborating on Richmond's ranking given that many have conceded now they will finish in the finals but do not think they will host a final. I said in the OP, the Tigers could do far, far better than that and could be top four in the rankings this week.

In terms of the biggest upsets, it's not how model takes into account things, but in terms of the biggest expected margin that turned out to be a loss was St Kilda v Dogs in round nine, Saints were expected to win by 56 points at the time. Port vs Sydney was the second biggest, Swans expected to win by 52 points and third was West Coast v Richmond, Eagles expected to win by 41 points.

 

[doppleganger]

You're right it should've been done earlier.

I can't comment on your rankings until you rationalise in real world values why you have chosen your particular metric. Here is a link which rationalises how the power rankings are formulated. If you have any further questions regarding those I will probably answer them next week when I have more time.

What was ur trigger to decide to now actually jump on a GF quinella?

Again, your link comments that the rankings are not designed to predict home and away games....yet you use them as a basis for 'discrepancies'...despite stating that they are not designed for home and away.

 

[Roby]

What was ur trigger to decide to now actually jump on a GF quinella?

Again, your link comments that the rankings are not designed to predict home and away games....yet you use them as a basis for 'discrepancies'...despite stating that they are not designed for home and away.

Of course they are it's an essential factor to equating performances.

 

[Roby]

Seven discrepancies found this week and eight bets put on this week, the most ever. Basically so that the rankings are showing to make profit again for the year.

Geelong v Fremantle – Handicap (+16.5) – $2 @ $1.91 Sportsbet

Richmond v St Kilda –Handicap (-28.5) – $2 @ $1.935 Pinnacle

West Coast v Essendon –Handicap (-4.0) – $2 @ $1.92 Betstar

Geelong v Fremantle – WIN– $1 @ $2.75 Centrebet (90 point or more money back offer)

Hawthorn v Brisbane –Handicap (-45.5) – $1.5 @ $1.92 Luxbet

Gold Coast v Adelaide–Handicap (-6.5) – $1.5 @ $1.92 Sportsbet

Port Adelaide v Collingwood–Handicap (+20) – $1.5 @ $1.92 Luxbet

GWS v North Melbourne –Handicap (-75.5) – $1 @ $1.92 Sportsbet

All the favourites are expected to come up this week except Geelong as they face the top ranked team this week. The closest margin predicted is for the Port v Collingwood game, Pies to win by 3 points.

Tips for round 14:

Dons
Swans
Crows
Pies
Freo
Dogs
Hawks
Roos
Tigers

(non-favourite in red)

Tips: 78/108

 

[mattis117]

Seven discrepancies found this week and eight bets put on this week, the most ever. Basically so that the rankings are showing to make profit again for the year.

Geelong v Fremantle – Handicap (+16.5) – $2 @ $1.91 Sportsbet

Richmond v St Kilda –Handicap (-28.5) – $2 @ $1.935 Pinnacle

West Coast v Essendon –Handicap (-4.0) – $2 @ $1.92 Betstar

Geelong v Fremantle – WIN– $1 @ $2.75 Centrebet (90 point or more money back offer)

Hawthorn v Brisbane –Handicap (-45.5) – $1.5 @ $1.92 Luxbet

Gold Coast v Adelaide–Handicap (-6.5) – $1.5 @ $1.92 Sportsbet

Port Adelaide v Collingwood–Handicap (+20) – $1.5 @ $1.92 Luxbet

GWS v North Melbourne –Handicap (-75.5) – $1 @ $1.92 Sportsbet

All the favourites are expected to come up this week except Geelong as they face the top ranked team this week. The closest margin predicted is for the Port v Collingwood game, Pies to win by 3 points.

Tips for round 14:

Dons
Swans
Crows
Pies
Freo
Dogs
Hawks
Roos
Tigers

(non-favourite in red)

Tips: 78/108

How do the Mundy and Barlow omissions change the outlook for the freo game?

If you were to name a Freo top 5 I am sure Mundy, Barlow, Pav and Sandi would be in or around it (McPharlin and Fyfe being 2 others) - if we still win without those guys it will be an epic victory.

 

[troywheatley]

Power Rankings Mid-Season Weightings

Just like last year, the mid-season weightings are up. This will show the reasoning behind the rankings positions as well show which teams have been penalised the most by umpires, had the worst injury run, who has been most and least competitive and who has been the most overrated/underrated.
.

Thanks for putting these up - as you say, they help give a clearer picture of why teams are ranked where they are.

(Actually, the last few pages of this thread seem to have a greater trend towards the analytics of ranking teams. It does make for a more informed discussion, but I can't help but feel that if those snide one-liners start to disappear then something will be lost ...)

 

[Simon_Nesbit]

Thanks for putting these up - as you say, they help give a clearer picture of why teams are ranked where they are.

(Actually, the last few pages of this thread seem to have a greater trend towards the analytics of ranking teams. It does make for a more informed discussion, but I can't help but feel that if those snide one-liners start to disappear then something will be lost ...)

Same thing happened last year. About mid-way through the season, the effect of prior-year performance was over-ruled by current, and the discrepencies between those teams with "easy" fixtures being up the top of the ladder were removed. The only trolls left were the bitter and twisted, and the anti-Hawks (who tbh have definitely deserved their #1 streak)

 

[troywheatley]

Same thing happened last year. About mid-way through the season, the effect of prior-year performance was over-ruled by current, and the discrepencies between those teams with "easy" fixtures being up the top of the ladder were removed. The only trolls left were the bitter and twisted, and the anti-Hawks (who tbh have definitely deserved their #1 streak)

I agree - I have no love for Hawthorn, but they were the #1 team who Sydney were good enough to outplay on that day in September.

 

[Roby]

How do the Mundy and Barlow omissions change the outlook for the freo game?

If you were to name a Freo top 5 I am sure Mundy, Barlow, Pav and Sandi would be in or around it (McPharlin and Fyfe being 2 others) - if we still win without those guys it will be an epic victory.

Mundy didn't play the week before either, so it had already been calculated, Barlow wasn't initially, as wasn't Spurr but later calculated. Nonetheless, it should've made no difference. Fremantle were poor because Geelong played their best four quarter game this year, while Fremantle were not able to match that all, in their least competitive of the year. They scored a 6 for competitiveness and before that game their lowest competitive score was 11, which was in their last loss against Hawthorn in round four.

Geelong had plenty of outs on the weekend, more than Freo, but despite not been in good form they somehow managed to find an extra level on the weekend when it was needed. Besides this match, it was the only game the rankings prediction was off otherwise it was pretty spot on this week.

Thanks for putting these up - as you say, they help give a clearer picture of why teams are ranked where they are.

(Actually, the last few pages of this thread seem to have a greater trend towards the analytics of ranking teams. It does make for a more informed discussion, but I can't help but feel that if those snide one-liners start to disappear then something will be lost ...)

It's not really an issue. Like any open forum you get a lot of different opinions, sometimes they are genuinely trying to further discussion, other times people have nothing better to do than annoy others. I say it's not an issue because you can spot quickly when this is the case, and ignoring is probably the best option. Otherwise it doesn't make any difference, the formula is the same, it doesn't change until it's reviewed at the end of the season. If people want to think I sit in a basement with a disheveled long beard tapping my fingers and concocting some interesting to way to re-rank teams in order to upset some unknowns on a internet forum, then I can only laugh.