Certified Legendary Thread Race for the flag, in squiggly lines

[Final Siren]

Round 18, 2014 (Week 1)​

Hard to see the squiggle this week through the smoke pouring off the tyres as St Kilda and Freo execute handbrake turns.

Safe to say the squiggle didn't see that result coming. I'm not sure how you would model a game like that. The theory I like is complacency + belief, i.e. Freo taking it easy colliding with a Saints side that had thrashed them late last year and so thought it could do it again. Typically sides respond to an unexpected loss by hardening up, e.g. Essendon repaying the Bulldogs for their Round 21 2000 loss by smashing them by 78 points in 2001. But Fremantle probably didn't see anything to avenge from their Round 23 2013 loss to the Saints, when Freo rested half their side.

North Melbourne are hiding behind Geelong, in shame I hope. Port keep sliding, with most of the year's gains gone now, as Melbourne get back on track.

And the Bulldogs deliver another quality performance, although they didn't win the match. That's three good games in a row for the Dogs. They're not getting much attention because it's included two close losses, but the squiggle sees a bit to like.

 

[SJ]

Who are you tipping between North and Geelong?

 

[threenewpadlocks]

Interesting to note that GWS, St. Kilda, Brisbane and Melbourne are very similarly ranked with varying levels of offence/defense.

GWS seems to be ahead slightly but the other three are neck and neck.

Unless I'm interpreting the squiggle wrong, am I?

 

[Final Siren]

Who are you tipping between North and Geelong?

The squiggle currently rates Geelong above North by a truly insignificant margin. It will tip North because it considers Geelong an interstate team and so awards North home advantage. But clearly it's a coin toss because who knows what North are doing.

 

[SJ]

The squiggle currently rates Geelong above North by a truly insignificant margin. It will tip North because it considers Geelong an interstate team and so awards North home advantage. But clearly it's a coin toss because who knows what North are doing.

Why does it consider Geelong an interstate team when it is not?

 

[Final Siren]

Interesting to note that GWS, St. Kilda, Brisbane and Melbourne are very similarly ranked with varying levels of offence/defense.

GWS seems to be ahead slightly but the other three are neck and neck.

Unless I'm interpreting the squiggle wrong, am I?

That's right, although I'd say all four are neck and neck at the moment, because the difference is so small. The amount that GWS are ahead of Melbourne is probably the equivalent of Jeremy Cameron getting a good night's sleep. But for the purposes of splitting hairs, right now it's GWS > St Kilda > Melbourne > Brisbane.

 

[Final Siren]

Why does it consider Geelong an interstate team when it is not?

Geelong is an oddity in that it's not an interstate team but it clearly enjoys an interstate-like home ground advantage at Kardinia. So when using a very simple model like ISTATE-91:12 that has to deem each team as interstate or not, it's more accurate when you treat Geelong as an interstate side.

In reality, over the last decade the Cats have had three levels of performance: Kardinia, Melbourne, and interstate. They've played a lot better than usual at Kardinia, better than usual in Melbourne, and worse than usual interstate. So to model that properly, you need a more sophisticated algorithm. (In fact it's even trickier than that, because when the Cats play a home game in Melbourne against another Melbourne-based team, they also play better than expected. This is a quirk I discovered by accident; there was a bug in one of my algorithms. But it holds up very reliably over the last decade.)

The squiggle is supposed to be simple enough to understand while providing a good model and visualization. So I don't use the more complicated algorithms here, even though they usually tip a little better. If they tipped a lot better, I would. But the difference isn't enough to justify it.

 

[Schauermann]

Geelong is an oddity in that it's not an interstate team but it clearly enjoys an interstate-like home ground advantage at Kardinia. So when using a very simple model like ISTATE-91:12 that has to deem each team as interstate or not, it's more accurate when you treat Geelong as an interstate side.

Actually an interstate side with less travel...

 

[wilbur_15]

The tips have updated to rnd 19 already on the site. Anyone write down/know how to see the tips for the 4 games this weekend?

 

[Inferno]

But clearly it's a coin toss because who knows what North are doing.

They'll do the opposite of whatever the squiggle thinks they'll do

 

[The Swans Blog]

Something is up with the Squiggles - Round 18 isn't completed, yet the tips for round 19 are up...


??

 

[Final Siren]

Something is up with the Squiggles - Round 18 isn't completed, yet the tips for round 19 are up...


??

Ah, thanks for letting me know. Datasource error. Fixed that.

 

[Final Siren]

The tips have updated to rnd 19 already on the site. Anyone write down/know how to see the tips for the 4 games this weekend?

ROUND 18 (Week 2) TIPS

West Coast 85 - 79 Richmond

Brisbane Lions 72 - 95 Gold Coast

Hawthorn 86 - 81 Sydney

Collingwood 83 - 76 Adelaide

​

Three games where the squiggle thinks the inferior team will win due to home state advantage.

 

[Dirty Bird]

ROUND 18 (Week 2) TIPS

West Coast 85 - 79 Richmond

Brisbane Lions 72 - 95 Gold Coast

Hawthorn 86 - 81 Sydney

Collingwood 83 - 76 Adelaide

​

Three games where the squiggle thinks the inferior team will win due to home state advantage.

I think that could go either 1/4 or 2/4, and maybe, just maybe 3/4 (but unlikely)

I agree on West Coast.
Disagree on Gold Coast
The other 2 are 50/50.

 

[KillaKanga]

I think that could go either 1/4 or 2/4, and maybe, just maybe 3/4 (but unlikely)

I agree on West Coast.
Disagree on Gold Coast
The other 2 are 50/50.

Could also go 4/4

 

[holybishop]

Does Brisbane have an advantage because they're playing at the Gabba? Or does the Squiggle assume it's a neutral venue for both.

 

[The Swert]

Three games where the squiggle thinks the inferior team will win due to home state advantage.

Hawthorn are inferior? Dangerous words.

Also are the Hawks currently forecast to win the Grand Final because of home ground advantage? Because the last 25 years haven't really shown any home ground advantage when it comes to grand finals.

 

[KissKiss]

Hawthorn are inferior? Dangerous words.

Also are the Hawks currently forecast to win the Grand Final because of home ground advantage? Because the last 25 years haven't really shown any home ground advantage when it comes to grand finals.

I think that Mr Squiggle (aka Final Siren) has stated this earlier that GF's are different (in terms of home ground advantage) and this algorithm isnt particularly successful for them. Other models have proved a little better for him

 

[estibador]

Lol @ the eagles being an inferior team to Richmond. The squiggle is wise.

 

[Final Siren]

Does Brisbane have an advantage because they're playing at the Gabba? Or does the Squiggle assume it's a neutral venue for both.

Neutral, as ISTATE-91:12 only looks at home state. (In a slightly warped universe where Geelong is its own state.)

 

[holybishop]

Neutral, as ISTATE-91:12 only looks at home state. (In a slightly warped universe where Geelong is its own state.)

Cant change it so the Gold Coast is their own state like Geelong?

 

[Roobs321]

Looks like it should be a good, close weekend of footy.

 

[Final Siren]

Hawthorn are inferior? Dangerous words.

Also are the Hawks currently forecast to win the Grand Final because of home ground advantage? Because the last 25 years haven't really shown any home ground advantage when it comes to grand finals.

More on Grand Finals later. I still need to hone some stats. But home state advantage does seem to be less of an advantage in Grand Finals... in fact there's some evidence the opposite is true.

But it's a slippery area because there's far less data. When I'm calibrating a model for general footy tipping, I can use data from 3,571 games since 1994. When I'm looking at Grand Finals over the same period, there are only 21 games, which makes it easier to imagine patterns that aren't there.

What I know for sure right now is that ISTATE-91:12 has been a lot worse at tipping Grand Finals than regular matches over the last 20 years. And it's been especially bad at tipping Grand Finals involving one Victorian team and one non-Victorian team. It would have done a lot better if it assumed that there was a home state disadvantage.

Another interesting thing is you can tip all finals except the Grand Final very accurately just by picking the home team. You don't need to know anything about football at all. Because of course the top teams are given home finals games, and they very often win. But then the GF rolls around and that evaporates.

 

[Final Siren]

Cant change it so the Gold Coast is their own state like Geelong?

There isn't enough data yet on whether that's justified... I'm not sure how many games the Suns have played in Brisbane but it wouldn't be many. Whereas with Geelong there are hundreds of matches to compare.

It makes intuitive sense, but I don't like adding variables on hunches; they have to prove themselves.

 

[Simon_Nesbit]

Another interesting thing is you can tip all finals except the Grand Final very accurately just by picking the home team. You don't need to know anything about football at all. Because of course the top teams are given home finals games, and they very often win. But then the GF rolls around and that evaporates.

Would it be considered too Roby-esque to conclude that this suggests the Grand Final is not a good way of determining the best team of the year?