Certified Legendary Thread Race for the flag, in squiggly lines

[Dr Tigris]

Just because something looks like a tool, doesn't mean it has a purpose.

The visualization of the flagpole is misleading and potentially incorrect (it's success rate isn't great before finals). Given it says it is in 'beta' I think it's open to criticism.

Why???

Just why?

The squiggle is a simple visualization of an algorithm designed to predict possible winners in football matches. It tends to provide a useful insight. That's all it is. It isn't aimed at winning the Nobel, or taught as theory at university.

 

[Illinois Nazi]

The flagpole is really just showing which team is playing the sort of footy that would likely win them a grand final if they played one tomorrow.

It doesn't have anything to do with whether or not they are likely to make it to a grand final in a month's time.

 

[kaiserchief13]

I stand corrected

1. Hawthorn 6-2 (3-1 v top 4)
2. Fremantle 6-4 (2-2 v top 4)
3. W Bulldogs 5-3 (2-3 v top 4)
4. West Coast 4-6 (2-2 v top 4)
5. Richmond 3-5 (3-2 v top 4)
6. North Melbourne 3-5 (2-2 v top 4)
7. Sydney 3-5 (1-3 v top 4)
8. Adelaide 3-5 (1-4 v top 4)

West Coast 5-5
Hawthorn - 1 loss
Fremantle - 1 loss 1 win
Bulldogs - 1 loss, 1 win
Richmond - 1 win
North - 1 loss
Sydney - 1 win
Adelaide - 1 loss, 1 win

 

[Final Siren]

[...] But even as a winner-takes-all model it isn't very successful. I imagine it was unlikely that flagpole picked Sydney 2005, Hawthorn 2008, Collingwood 2010, Geelong 2011 (before round 23), and Sydney 2012 as the clear premiership favourites in the latter rounds of those seasons. And I'm not sure where it rated Geelong against St Kilda in 2009 at round 22.

So at predicting the ultimate winner, it has no better than 50% success rate over the past decade.
[...]

It would be amazing if it tipped all those. You would basically have to stop watching football, because there would be no mystery left.

Picking premiers a month out is hard! The minor premier has only won four flags in the last decade, and six in the last 20 years. That's 30%! If you can pick premiers at 50%, you're doing great.

Anyway, in those years you mention:

• 2005: Definitely a bad year for prediction in general and flagpole especially, since it favours attacking teams, which Sydney were not. The Swans also seem to have used divine intervention to win at least one final, which is hard to model. They were #5 on flagpole after the home & away season and never higher than #4.
• 2008: Did all right: The Hawks were #2 for 16 weeks, including the last ten. They were never rated ahead of the Cats, though, that's for sure. The squiggle still thinks if you played that game again, Geelong would win.
  
• 2009: A close year: Geelong were rated #1 for 18 weeks, but were caught by the other four top teams near the end. The Cats were #2 going into prelims and 1st going into the GF... but there is a great quirk of the kind you are pointing out, where Adelaide remained the #1 team in prelim week even after being eliminated! And of course they were never marked down for finishing 5th, when that may have been fatal to their season, since they lost a semi-final interstate by 5 points.
  
• 2010: Did well: Collingwood were #1 on flagpole for seven weeks, including the last three. They were #1 or #2 the whole year after Round 5.
• 2011: Did well: Geelong were #2 the entire year until Round 23, when they moved to #1 and stayed there.
  
• 2012: Pretty poor, since it was Sydney being defensive again. They were mostly #6, up to #3 at the end of the home and away season and then #2 during finals, but always a lot lower than the Hawks.

 

[nameless]

Have 2 things been mentioned yet?

1) How Adelaide being 3rd on Flagpole lose to Dogs who are 6/7th?
2) Will next round just be written off due to so many teams resting players?

 

[SJ]

A lot of upsets in grand finals in the past decade means almost any sensible model would have been beaten.

 

[Hawkk]

You need to check your facts, some of those numbers are not correct.

West Coast are 5-5 against the top 8
Richmond are 3-5 against the top 8

Richmond and North play against each other in R23.

I notice you didnt read my post two posts after (where I replaced the Richmond numbers)

Good input but

 

[Hawkk]

West Coast 5-5
Hawthorn - 1 loss
Fremantle - 1 loss 1 win
Bulldogs - 1 loss, 1 win
Richmond - 1 win
North - 1 loss
Sydney - 1 win
Adelaide - 1 loss, 1 win

As above...

1. Hawthorn 6-2 (3-1 v top 4)
2. Fremantle 6-4 (2-2 v top 4)
3. W Bulldogs 5-3 (2-3 v top 4)
4. West Coast 5-5 (2-2 v top 4)
5. Richmond 3-5 (3-2 v top 4)
6. North Melbourne 3-5 (2-2 v top 4)
7. Sydney 3-5 (1-3 v top 4)
8. Adelaide 3-5 (1-4 v top 4)

 

[Bont2Bruce]

I notice you didnt read my post two posts after (where I replaced the Richmond numbers)

Good input but

Why not edit the incorrect post? Too lazy? Don't care?

I started a thread about this days before you posted your incorrect information, or any correction.

Questionable input from you.

 

[Final Siren]

Have 2 things been mentioned yet?

1) How Adelaide being 3rd on Flagpole lose to Dogs who are 6/7th?

Adelaide's form is rated above the Dogs both by flagpole and by the regular squiggle. But the predictor is tipping the Dogs to beat Adelaide in a match played in Melbourne because it uses the regular squiggle algorithm, ISTATE-91:12, which awards 12 points for home game advantage in games between interstate sides.

Historically, home sides overwhelmingly win finals (except the GF), even when there's no home state advantage.

2) Will next round just be written off due to so many teams resting players?

Not by the squiggle, which has no idea who's resting whom. Every point scored in every match counts.

In 2013, Fremantle rested half its team in R23, lost 112-41 to St. Kilda, then upset Geelong at Kardinia 87-72 in the first week of finals. You can see this as a hairpin turn:

 

[richoatthedisco]

Adelaide's form is rated above the Dogs both by flagpole and by the regular squiggle. But the predictor is tipping the Dogs to beat Adelaide in a match played in Melbourne because it uses the regular squiggle algorithm, ISTATE-91:12, which awards 12 points for home game advantage in games between interstate sides.

​

Does it take no notice of Richmond's excellent interstate record? 13-6 in the last 3 years.

 

[Suntory Boss Coffee]

Does it take no notice of Richmond's excellent interstate record? 13-6 in the last 3 years.

You simply just can't add "Richmond are good interstate". You need to find out how much numerically they are better interstate. And 19 pieces of data isn't enough

 

[sproketman]

The flagpole is really just showing which team is playing the sort of footy that would likely win them a grand final if they played one tomorrow.

It doesn't have anything to do with whether or not they are likely to make it to a grand final in a month's time.

But the flag isn't played tomorrow. Hence it is useless and misleading.

 

[Paracleet]

But the flag isn't played tomorrow. Hence it is useless and misleading.

Also if we take the combined probabilities for teams placed 5 to 8 by the 'flag', at the conclusion of the final round since 2000, of 0.x out 1 and compare it to the actual results, I.E. 0, we see there is an issue with it. It's useless without some scaling formula applied to the actual position on the ladder.

 

[wahawk]

But the flag isn't played tomorrow. Hence it is useless and misleading.

Well don't f##kin look at it then

 

[chazwazza]

Then it's not a hammer

It's a clammer

 

[Final Siren]

Here is basically why the squiggle is down on Freo: against the same opponent, Fremantle's results have been consistently poorer than other top teams.

For example, compared to West Coast, the Eagles have done better against 13 opponents, while Freo have done better against 4. (If you exclude the close ones where the difference is less than two goals, it's 11-1 in West Coast's favour.)

Opponent | vs Fremantle | vs West Coast | \ Adelaide | win by 11 (a) | win by 56 (h) | Freo +11 (avg)\ | | loss by 57 (a) \ Brisbane Lions | win by 36 (h) | win by 53 (a) | WC +17 \ Carlton | win by 42 (h) | win by 69 (h) | WC +27 \ Collingwood | win by 7 (h) | win by 31 (a) | WC +24 \ Essendon | win by 28 (h) | win by 50 (h) | WC +22 \ Geelong | win by 44 (a) | win by 56 (h) | WC +12 \ Gold Coast | win by 7 (a) | win by 92 (h) | WC +85 \ Greater Western Sydney | win by 21 (h) | win by 87 (h) | WC +66 \ Hawthorn | loss by 72 (a) | loss by 14 (h) | WC +58 \ Melbourne | win by 68 (a) | \ | win by 54 (h) | win by 54 (a) | Freo +7 (avg) \ North Melbourne | win by 73 (h) | \ | loss by 11 (a) | loss by 10 (a) | Freo +41 (avg) \ Port Adelaide | win by 7 (h) | win by 10 (a) | WC +3 \ Richmond | loss by 27 (h) | win by 20 (a) | WC +36 (avg)\ | win by 4 (a) | \ St Kilda | win by 37 (a) | win by 43 (a) | WC +6 \ Sydney | win by 14 (h) | win by 52 (h) | WC +40 \ Western Bulldogs | win by 13 (a) | loss by 10 (a) | WC +15 (avg) \ | | win by 67 (h) \ Derby (Freo vs WC) | win by 30 | loss by 30 | Freo +6 (avg)\ | loss by 24 | win by 24

As well as this, Fremantle's best games were all months ago, in the first 8 rounds.

Since the squiggle only looks at final scorelines and venue, it is never going to rate a team highly that has Fremantle's numbers.

 

[nurries]

Here is basically why the squiggle is down on Freo: against the same opponent, Fremantle's results have been consistently poorer than other top teams.

For example, compared to West Coast, the Eagles have done better against 13 opponents, while Freo have done better against 4. (If you exclude the close ones where the difference is less than two goals, it's 11-1 in West Coast's favour.)

Opponent | vs Fremantle | vs West Coast | \ Adelaide | win by 11 (a) | win by 56 (h) | Freo +11 (avg)\ | | loss by 57 (a) \ Brisbane Lions | win by 36 (h) | win by 53 (a) | WC +17 \ Carlton | win by 42 (h) | win by 69 (h) | WC +27 \ Collingwood | win by 7 (h) | win by 31 (a) | WC +24 \ Essendon | win by 28 (h) | win by 50 (h) | WC +22 \ Geelong | win by 44 (a) | win by 56 (h) | WC +12 \ Gold Coast | win by 7 (a) | win by 92 (h) | WC +85 \ Greater Western Sydney | win by 21 (h) | win by 87 (h) | WC +66 \ Hawthorn | loss by 72 (a) | loss by 14 (h) | WC +58 \ Melbourne | win by 68 (a) | \ | win by 54 (h) | win by 54 (a) | Freo +7 (avg) \ North Melbourne | win by 73 (h) | \ | loss by 11 (a) | loss by 10 (a) | Freo +41 (avg) \ Port Adelaide | win by 7 (h) | win by 10 (a) | WC +3 \ Richmond | loss by 27 (h) | win by 20 (a) | WC +36 (avg)\ | win by 4 (a) | \ St Kilda | win by 37 (a) | win by 43 (a) | WC +6 \ Sydney | win by 14 (h) | win by 52 (h) | WC +40 \ Western Bulldogs | win by 13 (a) | loss by 10 (a) | WC +15 (avg) \ | | win by 67 (h) \ Derby (Freo vs WC) | win by 30 | loss by 30 | Freo +6 (avg)\ | loss by 24 | win by 24

As well as this, Fremantle's best games were all months ago, in the first 8 rounds.

Since the squiggle only looks at final scorelines and venue, it is never going to rate a team highly that has Fremantle's numbers.

Didn't the Eagles draw The Gold Coast?

 

[Final Siren]

Didn't the Eagles draw The Gold Coast?

Possibly.

 

[Pessimistic]

R23 has screwed the finals predictor over.

Expect eagles to be standouts after their game

 

[Biggy_Boy]

Squiggle has gone mad. Is predicting a Swans vs. WCE grand final... That's not happening.

 

[Paul Kersey]

Squiggle has gone mad. Is predicting a Swans vs. WCE grand final... That's not happening.

Yeah was wondering how the hell it got that.

 

[nameless]

Edit: wrong thread

 

[bigton08]

Squiggle has gone mad. Is predicting a Swans vs. WCE grand final... That's not happening.

I think it could happen... There's been injuries though which could be the key... Don't believe freo will be there or the hawks.. Syds been flying under the radar a bit... And to make it more interesting I think a team outside of the four will be there..I rate Richmond .. But it's unlikely.but not impossible..if it's a WCE vs Sydney I hope WC win.

 

[Big Cox 88]

May as well give us the cup.