Certified Legendary Thread Race for the flag, in squiggly lines

[Final Siren]

Round 15, 2015

Animated!

This week the attacking teams attacked and the defensive teams defended! The Hawks piled on 17 goals against Freo, with the Dockers managing only 6 in reply, while the Eagles put 22 past Adelaide. The Swans could only manage 10 goals against bottom-placed Brisbane, the same number as Richmond kicked against Carlton.

As a result, the Hawks and Eagles soared further into prime premiership territory, opening up a bigger gap to the rest.

It was also a good week for the Roos, who recovered well from last week's debacle against the Suns to record a solid win over the Cats.

And it was a great week for squiggle tipping! 9/9 in a tough round.

Horrible week for Fremantle. It was always going to be tough to beat the Hawks in Tassie, but the Dockers were non-competitive, and again couldn't score.

The Dockers would need to get quite a lot worse to miss the Top 4 from here, since they're 3 wins and percentage ahead of 5th, but the ladder predictor has tipped them out of the Top 2:

This also puts Hawthorn a single win away from the minor premiership.

Now may I introduce... Squiggle Flagpole!

Flagpole aims to solve a problem with finals tipping I discussed earlier, where the chart shows Hawthorn in the best spot, but the predictor tips them to miss the Grand Final.

This is a bit of a pickle, because on the one hand, it's correct to tip the home team in any given regular final: they win 78% of the time. But on the other, the numbers say the Hawks are likely to get there somehow, whether it's by unexpectedly finishing top 2, or upsetting someone along the way.

So like the ladder predictor ignores individual tips to use a probabilistic approach, flagpole ignores individual tips to look at form. It does this with a different algorithm that I'll describe below, which is good at rating flag teams.

In simple terms, the higher the flag on the flagpole, the higher the likelihood of winning the premiership. The poles are also in order of predicted H&A finishing positions, using the regular ladder predictor. So this flagpole is saying the Hawks are expected to finish 4th but have the best flag form (by far!).

Feedback please! I can generate flagpoles for any round of any year, so if you have some ideas for interesting points in time to look at, please let me know, and we can see if the flagpole was being sensible or not.

How it works: Keen squiggle-watchers will know that the squiggle has a quirk whereby teams can move rightwards a long way by holding an opposition to a very low score. For example, see Round 8 this year, when Richmond held Port Adelaide to 43 pts, and subsequently executed the biggest horizontal jump of the year.

This initially looks like a bug in the model, but it actually seems to detect a real signal -- the Tigers were 3-4 (10th) before that game and are now 9-5 (5th). So that kind of thing is worth paying attention to.

I have another algorithm that works the same way, but in reverse, so that instead of gaining a big boost from a great defensive effort, teams gain it from kicking a big score. And I also have an algorithm that does both at once. This one is called OFFDEF3.

I made OFFDEF3 years ago but mostly ignored it because it's not great at regular tipping. But when I recently began to look at flag tipping, I noticed OFFDEF3 does pretty well. In fact, it's (relatively) good at tipping premiers at any stage of the season. So this is flagpole: OFFDEF3's rating of each team.

In a nutshell, OFFDEF3:

• ignores home ground advantage
• seriously rates teams that kick surprisingly big scores
• seriously rates teams that hold their opposition to surprisingly low scores
• seriously rates performances against good teams, especially later in the year, because it develops a ton more inflation than the regular squiggle

It's probably worth noting that Flagpole is highly influenced by the fact that attacking teams have won a lot of flags in the last 20 years. So one reason it's rating the Hawks is that Hawthorn 2015 looks a lot like Hawthorn 2014 and Hawthorn 2013. If St. Kilda and Fremantle had won more flags, I would probably be using a different algorithm, and not rating attacking teams so much.

OFFDEF3's Grand Final tipping is 16-4 over the last 20 years, ignoring the draw. Its errors are:

• 1997: Tipped St. Kilda by 14; Adelaide won by 31
• 2004: Tipped Brisbane by 18; Port Adelaide won by 40
• 2008: Tipped Geelong by 42; Hawthorn won by 26
• 2012: Tipped Hawthorn by 32; Sydney won by 10

Live squiggle (including Flagpole) and other follies

 

[Crankyhawk]

Surprised that flagpole rates port highly - or is that from previous results in 2014?

 

[YippeeYiYeo]

Great post. Like the addition of the flag pole

A couple of things:
- Freo since the Richmond game, have just about done a complete 180. In terms of premiership gradient cut off line that you could draw between Roos 99 and Sydney 2005, the dockers are as far away now as they were at the beginning of the season.
- Tigers, Magpies, Saints and Eagles have been the big movers of 2015 in a positive way.
- Bombers and Port Adelaide are the biggest losers on the squiggle this year.
- Stoked to see the Eagles move into premiership winning territory, but the next 4 weeks will decide if thats where we belong.
- Hawthorn are still the team to beat.

 

[Oneiros]

Sure is roomy out there

 

[eaglespremiers]

Great post. Like the addition of the flag pole

A couple of things:
- Freo since the Richmond game, have just about done a complete 180. In terms of premiership gradient cut off line that you could draw between Roos 99 and Sydney 2005, the dockers are as far away now as they were at the beginning of the season.
- Tigers, Magpies, Saints and Eagles have been the big movers of 2015 in a positive way.
- Bombers and Port Adelaide are the biggest losers on the squiggle this year.
- Stoked to see the Eagles move into premiership winning territory, but the next 4 weeks will decide if thats where we belong.
- Hawthorn are still the team to beat.

Access All Areas tends to be in agreement on this:
http://www.afl.com.au/video/2015-07-13/sam-you-need-to-change

 

[The Falcon Strike]

Can we all agree we won't go back through this thread and point out all my failings against the squiggle...

 

[The King!]

Would be happy with that week one final

 

[Shane Hird]

Great work FS..

 

[Lensen]

Can you produce a Flagpole (and advise where on the squiggle they sat) for Adelaide this round in 2006?

 

[Rotayjay]

There are a whole heap of points in various years where it would be interesting to see the flagpole. Which sides were in premiership form but people didn't rate them that highly? Which sides were highly rated, but flagpole 'knew' they were unlikely to win the flag?

Final Siren, would you be able to post the flagpole at Round 12 or so for the last twenty seasons? Then we might see which teams were in premiership form mid-season but fell short, which ones were 'always going to be' premiers, and which ones were not in premiership form mid-season but went on to take the flag.

 

[cryptor]

Would be interested to see how big the difference was between Geelong and Hawthorn on the flag pole immediately prior to the 2008 GF being played.

 

[Alberton]

I wouldn't want to see the 2001 - 2003 flagpoles.

 

[chiwigi]

Surprised that flagpole rates port highly - or is that from previous results in 2014?

Don't you think we are the 5th most likely team to win a premiership?
I'd say we are, not this season but overall.

 

[Crankyhawk]

Don't you think we are the 5th most likely team to win a premiership?
I'd say we are, not this season but overall.

I'd have Richmond ahead.

 

[KaaL]

Can we all agree we won't go back through this thread and point out all my failings against the squiggle...

I'm sure someone bookmarked it that you would run naked if we made top 4?

 

[Happy_Neale]

I wouldn't want to see the 2001 - 2003 flagpoles.

I do.

 

[eaglespremiers]

I'm sure someone bookmarked it that you would run naked if we made top 4?

That was pantskyle

 

[pantskyle]

That was pantskyle

I'll do it anyway.

 

[Roby]

Could be a range of things. Mid-season form slump, 4 weeks in a row playing in the wet, heavy training load, several players struggling for form etc.

Port went through the same thing last year but lost too many games during their slump to finish top 4. They recaptured their form leading into the finals and went within a whisker of beating Hawthorn in a prelim (they should have won that game with significantly more scoring shots).

I expect (and hope) for Freo to capture the form they showed in the first part of the season and finish the season strong.

Interesting points.

Mid-season form slump and players struggling for form

Teams can have form slumps, premiership teams tend not to. Generally when they get beaten they don't get beaten by a lot and there is usually a good reason like: missing personnel, umpiring deciding the game. I've watched every Freo game this year, and in the last two months they have had a bad run with the umpires. Having said that they still should be scoring more points. Their average betting line expected margin is 27 points in the last eight games. Their aggregate differential is 6 points, their 3.5 goals off their expected performance in two months of football. It's not a coincidence anymore.

Freo's top starting midfield is arguably the best in the competition. Fyfe, Sandi, Mundy, Neale, Barlow (Crowley when his back), Hill. The problem it drops massively after that. There is not enough midfield depth and the reason why they don't get enough good entries into their forward 50. Ball movement is slow if they aren't bursting from a stoppage.

You can say not having Johnson their also hurts but that's only one player. Most teams have far more inujries. Freo have had the least injured of their best 22 by far of any team this season. No wonder their on top of the table but apart from getting a home final it will count for nothing if they find their best form.

What has been Freo's best game this year in terms of everything going well? Was it the North or Dees game where they won by 10+ goals? Or was it the games against Sydney or Eagles. Against Sydney they nearly lost, and against the Eagles they looked like what Port did against Hawthorn, a fluke one quarter.

Four weeks in a row playing wet

I have model that takes into account playing in different weather conditions, including whether it's just dewy or very wet. I have taken that into account for Freo, the only issue with that is it tends to impact teams and how they finish off games. Fremantle has lost their starting quarter in their last six games. That's something to do with form and tactics. Something is not right other than fatigue if you cannot start well.


Port went through the same last year

So where are port now? Isn't this what I am forewarning? Port have a luxury of time of their side with a team which is 2.5 years younger than Freo, and have an average of less than 100 games experience. Freo had only less than 10 games experience on average than the Hawks the weekend, and you have to take into account the Hawks have added games from finals in recent years. They've changed the list a fair bit despite making finals each year, with trades and youth.


Great expectations

You expect Freo to regain their early season form, which whether you agree or not, I believe was not that good at its peak. Certainly not as good of football as either the Hawks or the Eagles at their peak. I've seen even the Dogs or Richmond play better football at their absolute best this year. Even if they do it doesn't look like it can win a premiership and that's hoping you don't get any injuries, which till now you've been pretty fortunate with. I actually don't think you will even beat Carlton this week, now if that happens what will be the dialogue, response and assessment?

 

[MF]

Interesting to see the squiggle placing Collingwood around 10th in the Flagpole.

 

[spinynorman]

Interesting points.

Mid-season form slump and players struggling for form

Teams can have form slumps, premiership teams tend not to. Generally when they get beaten they don't get beaten by a lot and there is usually a good reason like: missing personnel, umpiring deciding the game. I've watched every Freo game this year, and in the last two months they have had a bad run with the umpires. Having said that they still should be scoring more points. Their average betting line expected margin is 27 points in the last eight games. Their aggregate differential is 6 points, their 3.5 goals off their expected performance in two months of football. It's not a coincidence anymore.

Freo's top starting midfield is arguably the best in the competition. Fyfe, Sandi, Mundy, Neale, Barlow (Crowley when his back), Hill. The problem it drops massively after that. There is not enough midfield depth and the reason why they don't get enough good entries into their forward 50. Ball movement is slow if they aren't bursting from a stoppage.

You can say not having Johnson their also hurts but that's only one player. Most teams have far more inujries. Freo have had the least injured of their best 22 by far of any team this season. No wonder their on top of the table but apart from getting a home final it will count for nothing if they find their best form.

What has been Freo's best game this year in terms of everything going well? Was it the North or Dees game where they won by 10+ goals? Or was it the games against Sydney or Eagles. Against Sydney they nearly lost, and against the Eagles they looked like what Port did against Hawthorn, a fluke one quarter.

Four weeks in a row playing wet

I have model that takes into account playing in different weather conditions, including whether it's just dewy or very wet. I have taken that into account for Freo, the only issue with that is it tends to impact teams and how they finish off games. Fremantle has lost their starting quarter in their last six games. That's something to do with form and tactics. Something is not right if you cannot start well.


Port went through the same last year

So where are port now? Isn't this what I am forewarning? Port have a luxury of time of their side with a team which is 2.5 years younger than Freo, and have an average of less than 100 games experience. Freo had only less than 10 games experience on average than the Hawks the weekend, and you have to take into account the Hawks have added games from finals in recent years. They've changed the list a fair bit despite making finals each year, with trades and youth.


Great expectations

You expect Freo to regain their early season form, which whether you agree or not, I believe was not that good at its peak. Certainly not as good of football as either the Hawks or the Eagles at their peak. I've seen even the Dogs or Richmond play better football at their absolute best this year. Even if they do it doesn't look like it can win a premiership and that's hoping you don't get any injuries, which till now you've been pretty fortunate with. I actually don't think you will even beat Carlton this week, now if that happens what will be the dialogue, response and assessment?

I was pretty much with you until that last line. Expecting Fremantle to not beat Carlton?! Pull the other one!

 

[Roby]

I was pretty much with you until that last line. Expecting Fremantle to not beat Carlton?! Pull the other one!

I used this method see pic below (which beats the betting market) and blue came up instead of purple.

If only it picked yellow back when you played them back in round 10.

 

[Final Siren]

Surprised that flagpole rates port highly - or is that from previous results in 2014?

Quite a bit is 2014, yes. Which I think is appropriate: you don't want to write off last year's top teams too quickly. And OFFDEF3 is much less accurate at predicting premiers without that previous year's data. If it only looks at 2015, Port would be much lower (12th) and Collingwood would be much higher (3rd). West Coast would be closer to Hawthorn but everything else is pretty similar.

 

[lusheslewis]

Does the flagpole take into account that Port probably won't make the top 8?

 

[old55]

Analysis is one thing. Data visualisation is another. Whatever you think of the Squiggle on the first (and I like it), Final Siren has a real knack at the latter. Tremendous stuff, thanks!