Certified Legendary Thread The Squiggle is back in 2023 (and other analytics)

I think there's enough evidence that human assessment is definitely worse than a model, unless the human gets lucky for a while or is quite exceptional. Even the experts, who spend all day applying their human brain to the task, usually perform worse, as evidenced by your local newspaper's tips.

But I'm sure a human assessment often LOOKS better! I've seen lots of this on BigFooty, where people post rankings that are approved at the time by the hive mind, but then when you look at them later they were wildly wrong -- and not just wrong in the way where everyone was wrong (like how no-one tipped the last three premiers before the start of the season), which is understandable, but in the way where everything they claimed to have special insight about turned out to be off. That's fun to read about, and attracts attention, but is kind of unforgivable for a forecaster, since you're taking common wisdom and making it worse.

On the flip side, there is also a kind of selective memory when people make big calls and get them right! At first, this attracts ridicule, but if it turns out to be correct, popular sentiment seems to switch around to, "Yeah, but was that really surprising?" Everything seems obvious in retrospect, once it's already happened. So people underestimate what a great call it was, and forget that not many others made it.

Anyway, I guess I'm saying that if you wanted to compare the accuracy of human & computer forecasters, you should do so with real evidence, not just try to remember who posted what when.
But then what will people argue about on my forum???
 

SterlingArcher

Brownlow Medallist
May 16, 2014
12,163
15,983
AFL Club
Carlton
Both state and ground matters. Since last year, Squiggle uses a "Familiarity" method of establishing home ground advantage, which means the more often you've played at this ground and in this state compared to your opposition, the more HGA you have.

In reality, HGA is probably the result of several things, including crowd noise, umpiring, travel, fatigue, and psychology, which are all challenging to measure. But ground familiarity seems to be a pretty good working proxy for those.

Theoretically if ground familiarity is important, does it look how often a team plays there or the sum of the players named?

I.e. Richmond would have been a lot less 'familiar' with no Rance (9 previous AO games), Cotchin (9), Martin (8), Houli (6), Riewoldt (9), Grigg (9), Short (5) for a total of 55 games - replaced by Menadue (4), Broad (2), Ross (0), Bolton (1), Balta (0), Baker (0), Stack (0) for a total of 7.

Or is that not how it's interpreted?
 
Apr 28, 2008
11,211
8,194
AFL Club
West Coast
Other Teams
Arsenal Kilmarnock
When was the last time Sydney were on bottom of the Squiggle rankings?
Probably as late as a round into 1995.

Started 1995 0-3, but their losses were honourable whilst Fitzroy and St Kilda were already getting flogged. They were the bottom 3 in 1994. Fitzroy were 3-2 in early 1994, so their squiggle would have tanked hard in early 1995. Fremantle going 0-2 in their inaugural season is also an interesting angle there.

Fitzroy beating Sydney at the SCG in Rnd 20 1994 is an interesting match, as Fitzroy were probably overtaking Sydney at the bottom of the barrel at that point.

I can't think of any other time over the last 25 years. Possibly winter 2002, as the Dockers, Eagles and Saints were more capable than 2001, whilst the Blues and Tigers were falling from a greater height. Bulldogs had a slow start after an ordinary year. But the Swans had a lot of close losses before Eade went, so they weren't easybeats. I suspect the Saints were always behind the Swans then though.
 
Last edited:
Sep 13, 2015
18,683
48,481
Hillary Step
AFL Club
Richmond
Other Teams
76ers
Probably as late as a round into 1995.

Started 1995 0-3, but their losses were honourable whilst Fitzroy and St Kilda were already getting flogged. They were the bottom 3 in 1994. Fitzroy were 3-2 in early 1994, so their squiggle would have tanked hard in early 1995.

Fitzroy beating Sydney at the SCG in Rnd 20 1994 is an interesting match, as Fitzroy were probably overtaking Sydney at the bottom of the barrel at that point.

I can't think of any over time over the last 25 years. Possibly winter 2002, as the Dockers, Eagles and Saints were more capable than 2001, whilst the Blues and Tigers were falling from a greater height. Bulldogs had a slow start after an ordinary year. But the Swans had a lot of close losses before Eade went, so they weren't easybeats.
I'm not gonna look at every round between then and now, but by eyeballing they did start lowest ranked in 95 (just), so that's a shout.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
Theoretically if ground familiarity is important, does it look how often a team plays there or the sum of the players named?

I.e. Richmond would have been a lot less 'familiar' with no Rance (9 previous AO games), Cotchin (9), Martin (8), Houli (6), Riewoldt (9), Grigg (9), Short (5) for a total of 55 games - replaced by Menadue (4), Broad (2), Ross (0), Bolton (1), Balta (0), Baker (0), Stack (0) for a total of 7.

Or is that not how it's interpreted?
The familiarity of individual players with the ground isn't considered, no.

My guess is that's not important, for two reasons.

Firstly, I don't think what matters is actual familiarity with the ground. There's some pretty good evidence that the #1 factor in home ground advantage is crowd passion, which influences umpiring decisions. But that's very difficult to measure, or to predict in advance, so people like me look for something they can measure, which correlates well with crowd passion. Ground/state familiarity seems to be a pretty good choice, because if a team plays somewhere much more often than their opponent, they practically always have the lion's share of crowd support, too.

This isn't perfect, of course. West Coast and Fremantle both play lots of games at Perth Stadium, for example, but West Coast reliably have louder, more passionate crowds, and likewise seem to have more home ground advantage there. But it's decent.

Secondly, I think even if you carefully tracked every player's individual familiarity with every ground, it would turn out to be a vanishingly small difference compared to simply treating the team as one entity. It's hard to find really significant factors that influence results, and even things that you'd think would be pretty obvious, like how the amount of rest a team has had between matches, are difficult to find evidence for in the stats. Richmond took over a team with less ground familiarity than usual to South Australia, for example, but performed better than expected, which runs counter to the idea that ground familiarity matters at all.
 

willydave

All Australian
Oct 25, 2002
901
316
Melbourne
AFL Club
Melbourne
Sneak preview: I'm starting to collect ladder predictions from the internet's best computer models and condense them into a single ladder. Let me know what you think!

https://squiggle.com.au/ladder/
I know we've (Demons) been rubbish so far, but some of the projected losses/draw for us are amusing. We are slow to get going this year due to fitness and we will take care of teams like the Suns when we hit them.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
I know we've (Demons) been rubbish so far, but some of the projected losses/draw for us are amusing. We are slow to get going this year due to fitness and we will take care of teams like the Suns when we hit them.
You well might, but boy, you can understand why people have jumped off. Melbourne's results have been horrendous. Even the R4 victory doesn't look great, since the Swans are also struggling, and the Demons won only because of a very unusual goalkicking distribution, where Sydney were unusually inaccurate and Melbourne were unusually accurate. If every player had kicked with normal accuracy, it's a comfortable Swans win and Melbourne are 0-4 (after 0-2 in JLT).

Squiggle is even tipping the Saints over Melbourne this week, although that's mainly because Squiggle thinks the Swans are really terrible, so doesn't rate the Dees' R4 win at all.
 
I know we've (Demons) been rubbish so far, but some of the projected losses/draw for us are amusing. We are slow to get going this year due to fitness and we will take care of teams like the Suns when we hit them.

Sure you will ;)
 

Mantastic

Premiership Player
Jun 22, 2011
3,996
5,060
AFL Club
Geelong
Other Teams
Man City, Cleveland Browns
I know we've (Demons) been rubbish so far, but some of the projected losses/draw for us are amusing. We are slow to get going this year due to fitness and we will take care of teams like the Suns when we hit them.

Yeah, the Dee's are great this year. Ha.
 

Isaac Cumming No 1

Cancelled
10k Posts
Mar 28, 2018
10,491
8,454
AFL Club
GWS
Some of these ‘upsets’ are wins by teams higher on the ladder.
No doubt some if it is pre-season expectations colliding with reality. It's hard to line up form in the early rounds as well.

Look at Freo's win yesterday for example. They were brilliant in round 1. They beat the Saints which wasn't rated at the time but might just stand up. They played well in the derby but were paying $6 yesterday.

Maybe algorythms just take time to catch up.
 
Sep 25, 2007
13,998
14,223
AFL Club
West Coast
No doubt some if it is pre-season expectations colliding with reality. It's hard to line up form in the early rounds as well.

Look at Freo's win yesterday for example. They were brilliant in round 1. They beat the Saints which wasn't rated at the time but might just stand up. They played well in the derby but were paying $6 yesterday.

Maybe algorythms just take time to catch up.

After GWS won at Geelong, $6 hardly seems overly generous, even after the fact for Freo.
 

Hobbes

Club Legend
Jul 20, 2006
1,846
2,551
Oxfordshire, UK
AFL Club
Adelaide
Other Teams
Glenelg
Brisbane v Collingwood +10
North Melbourne v Essendon +16
West Coast +39 v Port Adelaide
GWS +26 v Fremantle
Melbourne v St Kilda +3
Richmond +16 v Sydney
Western Bulldogs +17 v Carlton
Adelaide +14 v Gold Coast
Hawthorn v Geelong +21

6/9. Running total 24/45. I don't think anybody saw Port, Freo or Carlton coming.


1. Collingwood 19.9 (+3)
2. Geelong 19.1
3. West Coast 10.9 (-2)
4. GWS 9.8 (-1)
5. Fremantle 9.6 (+1)
6. Essendon 9.0 (-1)
7. Port Adelaide 7.8 (+6)
8. St Kilda 6.4
9. Adelaide 2.5 (+5)
10. Richmond 0.3
11. Hawthorn -2.8
12. Brisbane -6.2 (-5)
13. Carlton -9.1 (+5)
14. Melbourne -9.4 (-2)
15. Western Bulldogs -9.7 (-6)
16. Gold Coast -11.7 (-1)
17. Sydney -13.6
18. North Melbourne -16.9 (-1)

Tons of movement here as the pack in the middle starts to spread out, and West Coast and GWS fall back toward the field.

Richmond +10 v Melbourne
Essendon v Collingwood +11
Port Adelaide +31 v North Melbourne
Gold Coast v Brisbane +5
St Kilda +10 v Adelaide
Sydney v GWS +23
Fremantle +28 v Western Bulldogs
Hawthorn +6 v Carlton
Geelong +17 v West Coast
 
Back