AFL Power Rankings 2014

[Malibu#27]

Roby, why don't you publish the data you use to generate your umpiring decisions so we can see what assumptions you are using to base your rankings on.
Lets see if your judgement is as good as you think it is.

Also - how do the rankings look if you only take the last 2 seasons (not 3 as I think you use into account).


Finally, just reading that Carlton went up this week. Now no maths on earth can justify that, and it alone shows a fundamental flaw in your system.

 

[Schulzenfest]

Roby, why don't you publish the data you use to generate your umpiring decisions so we can see what assumptions you are using to base your rankings on.
Lets see if your judgement is as good as you think it is.

Don't be silly. We're missing Wingard, Monfries, Moore and Redden from our best 22 through injury and we have the worst free kick differential in the competition, but we're obviously the luckiest team and deserve to be seven spots lower on the Power Rankings ladder than the real ladder. I don't know how you could possibly question that.

 

[Lensen]

moore and redden are massive losses :/

 

[eth-dog]

moore and redden are massive losses :/

One of the best small forward combo in the game is a huge loss though.

 

[Fadge]

moore and redden are massive losses :/

Who are Moore and Redden? Doesn't Redden play for Brisbane?

 

[Lensen]

One of the best small forward combo in the game is a huge loss though.

sure, but it's not like two guns out for two games each (plus approx half a game in monfries' case) is a particularly unlucky occurrence

 

[Malibu#27]

Who are Moore and Redden? Doesn't Redden play for Brisbane?

Andrew Moore is #26 in the picture below taken from our victorious elimination final against some mob.

 

[Fadge]

sure, but it's not like two guns out for two games each (plus approx half a game in monfries' case) is a particularly unlucky occurrence

A bit like Monkey King quoting an injured Mitch Robinson as a loss for the Blues when he went off halfway through the last quarter and was right to play the following week.

 

[Lensen]

if anything, andrew moore's absence is a stroke of luck. since hinkley came on board, port are 8-9 with moore playing and 8-3 without

 

[Sellyme]

if anything, andrew moore's absence is a stroke of luck. since hinkley came on board, port are 8-9 with moore playing and 8-3 without

In the last five years both Sydney and Hawthorn have been more than twice as likely to lose a match if Lance Franklin is playing.

 

[Duck Farmer]

Despite losing to Melbourne (18th) this week, they actually move up the rankings, wowwee!

Just when I thought the best thread on BigFooty couldn't get any better

 

[Donakebab]

Oh mighty Roby, please share your infinite wisdom and inform us lowly peasants as to how Essendon are considered lucky with injuries?

 

[Astro Bomber]

Roby, why don't you publish the data you use to generate your umpiring decisions so we can see what assumptions you are using to base your rankings on.
Lets see if your judgement is as good as you think it is.

Also - how do the rankings look if you only take the last 2 seasons (not 3 as I think you use into account).


Finally, just reading that Carlton went up this week. Now no maths on earth can justify that, and it alone shows a fundamental flaw in your system.

yeah. Let's see your umpiring stats Roby. That's not revealing any of your IP.

 

[Roby]

How does this work? Wouldn't it make more sense for teams being defeated by better teams (expected result) to not surrender many points, whilst teams falling to worse teams (unexpected) to lose significant points?

How can a team losing an effective 'walkover' result be taken as not such a bad thing?

You would think so, that's essentially how many more simple models work but essentially why they are always flawed and always fail.

Mathematicians like to deal with simple absolutes like the one you just mentioned because it appears logical, the problem with that is that makes a incorrect assumption, that is that players are robots and perform to an certain expected output each time.

Before I even posted the first power rankings four years ago, I explained that I had success in modelling for finals outcomes and in betting over the years and decided to share it. I had tried many systems and realised this was the best system.

Rather than finding a system which "made sense" I just kept remodeling one that became more and more accurate.

The findings for me a quite simple: Teams and players play to a certain intensity depending on the opponent, the time of year, and a whole host of other factors.

In this case, Melbourne was and is seen by other teams as a minimal effort game. Carlton believed that they were not too far off a premiership, in fact, around this time in 2012, they were not just considered a premiership favorite, they were the premiership favourite according to the betting market. It's the reason they went and got Thomas and Malthouse, yes?

An outsider just says last weekend, 'well why didn't they just try harder, didn't they know what was at stake?', but the club and players could have a completely different opinion of where they are at and may have thought they had bigger fish to fry, thinking they could cruise to victory.

Now they can't really put half baked performances anymore, they need to go full throttle or risk undoing all the hard work they have done in previous years. Essentially what happens is Carlton becomes like Melbourne, and even if they start winning games it will because other teams will be doing the same thing to them, not trying as hard to beat them and then creating a false sense of their true ability or premiership potential. This is what I reported after the first round Power Rankings:

Carlton (15th) plummet to their lowest position in rankings history. Time for rebuild.

This is what has happened to Melbourne, no matter how hard they tried in recent years, other teams have never shown them a competitive true reflection of themselves so that Melbourne can put their gameplan on the ultimate test week after week.

On the other hand, teams like Geelong and Hawthorn have been seen as the ultimate benchmark, it makes it real easy for Scott and Clarkson to test their gameplans and teams constantly to the best litmus tests, further improving on it with each week.

History shows that many teams that do well in finals or win premierships can lose walkover games/deadrubbers, especially late in the season, and therefore not matter, but when they lose to finals competitors there is a significant correlation with their impact in September.

Time to put your money where your mouth is, Roby. You'd get astronomical odds on a Crows premiership, and yet according to you they're one of only 4 teams who can win the flag (despite having only one win this season against the team that was widely tipped pre-season to get the spoon). That's value if I've ever seen it. Put a big chunk of cash on it, and post proof.

Prediction model is good for that but injuries need to be quantified. Seems worthwhile to have a look at.

I find that stat really interesting, but I don't reckon it's a forgone conclusion. Can guarantee Adelaide will not be one of the last 4 teams in September, and anyone that makes a prelim has a chance of winning it on their day.

You think the Swans could've won the grand final last year if Fremantle had forfeited?

Roby I might have missed it - what was your starting bankroll re punting?

$6 on weekly bets and $12 including long term betting.

Then an additional $3.85 the first two weeks and $9.85 is the most taken out for weekly bets. An additional $9 currently for the year long terms bets totaling $18.85 for weekly and long term.

If either Hawthorn wins the premiership, or it's a Hawthorn/Fremantle grand final, the long terms bets will profit, with then surplus for Suns making the finals, Adelaide top 4, Eagles missing the finals.

Fremantle could also win the premiership but then it would need at least any of the other long terms bets in order to profit.

I'd love to be in the 8, but sorry Roby, we are a bottom 4 quality side right now. The only reason the result against Collingwood looked semi respectable is because they didn't give a s**t in the last and we piled on 7 goals. Essendon and Port Adelaide are a genuinely good teams and deserve to be higher. Adelaide are not in the best 4 teams of the comp. Sydney also too high.

It's not just about right now though, is it.

Oh mighty Roby, please share your infinite wisdom and inform us lowly peasants as to how Essendon are considered lucky with injuries?

Round 1: Goldstein plays with one arm for the whole game, Cunnington (he's an important player for North right) was concussed and missing for a quarter.

Round 2: Essendon played a very depleted Hawthorn team, missing key position players, and also in the same department/s.

Since then they rank as the 8th unluckiest on the TIT (Total Injury Table).

 

[Final Siren]

That is truly fascinating, and I have to say Roby I have never encountered a mathematical model that allows for teams not trying.

It makes a certain amount of intuitive sense, although I do wonder how you determine exactly how hard a team is likely to try. Is this a subjective measure (e.g. input roby's feels on a scale of 1 to 10 for each match) or an objective one (e.g. pop out a number derived from ladder differential + round number + ...)?

I also wonder how you were able to "just keep remodeling" a system that requires you to carefully watch each match. Are you still talking about only 2011-2013 here, or have you been able to compare some part of your model against results from previous seasons?

At first, I assumed that "Carlton rises in the power rankings because they only lost to a bad team, not to a good one" must be a typo, but this is a lot more interesting.

 

[ManWithNoName]

You would think so, that's essentially how many more simple models work but essentially why they are always flawed and always fail.

Mathematicians like to deal with simple absolutes like the one you just mentioned because it appears logical, the problem with that is that makes a incorrect assumption, that is that players are robots and perform to an certain expected output each time.

Before I even posted the first power rankings four years ago, I explained that I had success in modelling for finals outcomes and in betting over the years and decided to share it. I had tried many systems and realised this was the best system.

Rather than finding a system which "made sense" I just kept remodeling one that became more and more accurate.

The findings for me a quite simple: Teams and players play to a certain intensity depending on the opponent, the time of year, and a whole host of other factors.

In this case, Melbourne was and is seen by other teams as a minimal effort game. Carlton believed that they were not too far off a premiership, in fact, around this time in 2012, they were not just considered a premiership favorite, they were the premiership favourite according to the betting market. It's the reason they went and got Thomas and Malthouse, yes?

An outsider just says last weekend, 'well why didn't they just try harder, didn't they know what was at stake?', but the club and players could have a completely different opinion of where they are at and may have thought they had bigger fish to fry, thinking they could cruise to victory.

Now they can't really put half baked performances anymore, they need to go full throttle or risk undoing all the hard work they have done in previous years. Essentially what happens is Carlton becomes like Melbourne, and even if they start winning games it will because other teams will be doing the same thing to them, not trying as hard to beat them and then creating a false sense of their true ability or premiership potential. This is what I reported after the first round Power Rankings:




This is what has happened to Melbourne, no matter how hard they tried in recent years, other teams have never shown them a competitive true reflection of themselves so that Melbourne can put their gameplan on the ultimate test week after week.

On the other hand, teams like Geelong and Hawthorn have been seen as the ultimate benchmark, it makes it real easy for Scott and Clarkson to test their gameplans and teams constantly to the best litmus tests, further improving on it with each week.

History shows that many teams that do well in finals or win premierships can lose walkover games/deadrubbers, especially late in the season, and therefore not matter, but when they lose to finals competitors there is a significant correlation with their impact in September.





Prediction model is good for that but injuries need to be quantified. Seems worthwhile to have a look at.





You think the Swans could've won the grand final last year if Fremantle had forfeited?





$6 on weekly bets and $12 including long term betting.

Then an additional $3.85 the first two weeks and $9.85 is the most taken out for weekly bets. An additional $9 currently for the year long terms bets totaling $18.85 for weekly and long term.

If either Hawthorn wins the premiership, or it's a Hawthorn/Fremantle grand final, the long terms bets will profit, with then surplus for Suns making the finals, Adelaide top 4, Eagles missing the finals.

Fremantle could also win the premiership but then it would need at least any of the other long terms bets in order to profit.





It's not just about right now though, is it.





Round 1: Goldstein plays with one arm for the whole game, Cunnington (he's an important player for North right) was concussed and missing for a quarter.

Round 2: Essendon played a very depleted Hawthorn team, missing key position players, and also in the same department/s.

Since then they rank as the 8th unluckiest on the TIT (Total Injury Table).

So in other words: "No"?

 

[ManWithNoName]

Hypothesis: Roby is in fact Sir Humphrey Appleby.

 

[lamaros]

That is truly fascinating, and I have to say Roby I have never encountered a mathematical model that allows for teams not trying.

It makes a certain amount of intuitive sense, although I do wonder how you determine exactly how hard a team is likely to try. Is this a subjective measure (e.g. input roby's feels on a scale of 1 to 10 for each match) or an objective one (e.g. pop out a number derived from ladder differential + round number + ...)?

I also wonder how you were able to "just keep remodeling" a system that requires you to carefully watch each match. Are you still talking about only 2011-2013 here, or have you been able to compare some part of your model against results from previous seasons?

At first, I assumed that "Carlton rises in the power rankings because they only lost to a bad team, not to a good one" must be a typo, but this is a lot more interesting.

The basic rationale has some merit (that teams modify their performance based on expectation), its actually the reason I don't often bet lines, however the implementation is a load of rubbish.

 

[Roby]

That is truly fascinating, and I have to say Roby I have never encountered a mathematical model that allows for teams not trying.

It makes a certain amount of intuitive sense, although I do wonder how you determine exactly how hard a team is likely to try. Is this a subjective measure (e.g. input roby's feels on a scale of 1 to 10 for each match) or an objective one (e.g. pop out a number derived from ladder differential + round number + ...)?

I also wonder how you were able to "just keep remodeling" a system that requires you to carefully watch each match. Are you still talking about only 2011-2013 here, or have you been able to compare some part of your model against results from previous seasons?

At first, I assumed that "Carlton rises in the power rankings because they only lost to a bad team, not to a good one" must be a typo, but this is a lot more interesting.

It's not derived from ladder position, it's purely from the power rankings. Is this correct? Do teams rate other teams similar to these rankings? Difficult, maybe impossible to quantify, but measures can be put in place to track if you know what you're looking for. Just takes a lot of time if one cares enough.

I can compare back all the way to 1897. As stated each and as far back as 2011 I've researched and based the power rankings model on VFL/AFL records.

 

[SJ]

It's not derived from ladder position, it's purely from the power rankings. Is this correct? Do teams rate other teams similar to these rankings? Difficult, maybe impossible to quantify, but measures can be put in place to track if you know what you're looking for. Just takes a lot of time if one cares enough.

I can compare back all the way to 1897. As stated each and as far back as 2011 I've researched and based the power rankings model on VFL/AFL records.

How do you go factoring umpiring bias in to 1897 results?

 

[Sellyme]

How do you go factoring umpiring bias in to 1897 results?

He watched every single game, geez, don't you read?

 

[Crankyhawk]

Richmond $2.35 LOSS
Carlton $1.13 LOSS
Port $1.13 WIN
GWS $2.20 LOSS
GC $4.00 LOSS
Geelong $1.30 WIN
Sydney $1.38 LOSS
St Kilda $2.20 (surely that's unders) LOSS
Freo $1.80 WIN

Was $2.24 ahead on $23 total; now ($1.13 + 1.30 + 1.80 = $4.23) so $2.53 running loss on $32 wagered cumulatively.

This week
Brisbane $4.35
Collingwood $1.83
Sydney $2.20 (what? unders)
Essendon $1.09
WCE $1.87 (unders again)
Adelaide $1.19
Melb $3.30
Dogs $1.55
Cats $2.50

 

[Roby]

Power Rankings betting tips for Round 5 - Another record betting punt

Last week was the most the rankings have ever staked in one week, and the exact amount will be staked amount this week. Of course if all the bets were to fail it would take more than two weeks at this betting rate before we need to initiate additional funds but given that the rankings are getting closer to the line in 77% of games in the last fortnight (66% for the year) it is unlikely that is to happen. Of course the kings may have the final say.

Of course there is the "negative & positive continuity theory" which basically means that at some point the market will adjust to the true value of teams, (which should be in line with these rankings) and then the eventually drop off and improvement of teams will inevitably mean a weekly loss will be posted. At this point if the rankings make profit this week it will be nine weeks in a row and I'm guessing that could be some kind of world record.

Adelaide vs GWS (-35.5) - $3.5 @ $2 Topsport

Brisbane vs Richmond (-25.5) - $2.5 @ $1.943 Matchbook

Essendon vs St Kilda (+46.5) - $2.5 @ $1.91 Centrebet

Hawthorn vs Geelong (-10.5) - $2 @ $1.925 Matchbook

West Coast vs Port Adelaide (+2.5) - $1.5 @ $1.91 Centrebet

Collingwood vs Kangaroos (+2.5) - $1.5 @ $2 Topsport

Gold Coast vs Melbourne (+21.5) - $1 @ $1.92 Sportsbet

Bulldogs vs Carlton (-8.5) - $0.5 @ $1.92 Luxbet

Sydney vs Fremantle (-5.5) - $0.5 @ $1.901 Pinnacle

Current fund: $35.77
Total staked for 2014: $18.85
Last week's betting tips. Here. And here.

Power Rankings tips for Round 5

Two more non-favourites are expected to come up this week, so far the rankings have 3/7 in non-favourites getting up and 2/2 in even money games.
Tigers
Roos
Freo
Port
Dons
Crows
Suns
Dogs
Hawks


2014 tip tally: 25/36

 

[ManWithNoName]

Power Rankings betting tips for Round 5 - Another record betting punt

Last week was the most the rankings have ever staked in one week, and the exact amount will be staked amount this week. Of course if all the bets were to fail it would take more than two weeks at this betting rate before we need to initiate additional funds but given that the rankings are getting closer to the line in 77% of games in the last fortnight (66% for the year) it is unlikely that is to happen. Of course the kings may have the final say.

Of course there is the "negative & positive continuity theory" which basically means that at some point the market will adjust to the true value of teams, (which should be in line with these rankings) and then the eventually drop off and improvement of teams will inevitably mean a weekly loss will be posted. At this point if the rankings make profit this week it will be nine weeks in a row and I'm guessing that could be some kind of world record.

Adelaide vs GWS (-35.5) - $3.5 @ $2 Topsport

Brisbane vs Richmond (-25.5) - $2.5 @ $1.943 Matchbook

Essendon vs St Kilda (+46.5) - $2.5 @ $1.91 Centrebet

Hawthorn vs Geelong (-10.5) - $2 @ $1.925 Matchbook

West Coast vs Port Adelaide (+2.5) - $1.5 @ $1.91 Centrebet

Collingwood vs Kangaroos (+2.5) - $1.5 @ $2 Topsport

Gold Coast vs Melbourne (+21.5) - $1 @ $1.92 Sportsbet

Bulldogs vs Carlton (-8.5) - $0.5 @ $1.92 Luxbet

Sydney vs Fremantle (-5.5) - $0.5 @ $1.901 Pinnacle

Current fund: $35.77
Total staked for 2014: $18.85
Last week's betting tips. Here. And here.

Power Rankings tips for Round 5

Two more non-favourites are expected to come up this week, so far the rankings have 3/7 in non-favourites getting up and 2/2 in even money games.
Tigers
Roos
Freo
Port
Dons
Crows
Suns
Dogs
Hawks


2014 tip tally: 25/36

So you have one more tip than I do, and I forgot to put my tips in last week.

 

[Sellyme]

2014 tip tally: 25/36

I don't know what most tipsters are on at the moment, but in the BAU forum tipping competition, the two people who are actual avid followers of the sport (myself being one) are on 25 as well (I'm actually on 26, changed my tip on Haw v Freo but forgot to update it across the site), and a Thai with very little football knowledge is on 27. Your tip score doesn't seem very impressive considering it's equaled or beaten by people just going off a hunch.

Additionally, your "world record" of 9 consecutive "weeks" is beaten consistently by anyone who does nothing but bet against Melbourne each week. Would've been, what, 12 rounds straight before Saturday? And while you're counting round 1 as two weeks, why not count all the weeks October-February too? They're just as valid for your arbitrary on-the-spot definitions.