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

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
Squiggle hates Freo, as I understand it's because of our crap percentage and winning games we shouldn't etc last year, then we lost Cerra so that can't be great for ins/outs


Does it take into account our injury list this year compared to previous years or will that be in the weekly ins/out calcs

Also our accuracy was dogshit last year did it factor that in when it rates us so poorly? does squiggle just expect us to kick like 8.16 every week now or is it a clean slate each year and it just rates us as low scoring?
Most of the models up and running so far are pretty grim on Freo:

Screenshot from 2022-03-01 10-31-20.png


This is mostly because of, as you say, crap percentage and winning close games, but also because of how the Dockers fell off towards the end of the year.

Percentage is often a better indicator of underlying team performance than premiership points, because it's harder for a bit of luck to shift the number one way or the other. It only takes three lucky kicks at the right moments to earn 12 premiership points, but to significantly move your percentage, you need things to go very well (or badly) over a much more sustained period.

So it's not promising that Freo's 2021 percentage (86%) was only significantly better than two teams: Gold Coast (77%) and North Melbourne (70%), while being about the same as four other teams who all finished below them: Carlton (89%), Hawthorn (85%), Adelaide (82%), and Collingwood (86%).

The injury list is considered by Squiggle, but not, I think, any of those other models. Squiggle sees it like this:

Code:
FREMANTLE
          Jordan Clark     +57.6 
           Will Brodie     +26.6 
           Leno Thomas      -0.0 
           Stefan Giro     -22.7 
          Stephen Hill     -54.9 
           Tobe Watson     -58.7 
        Connor Blakely     -65.3 
           Reece Conca    -100.6 
          Brett Bewley    -104.7 
          Taylin Duman    -134.5 *
      Mitchell Crowden    -165.8 *
            Adam Cerra    -350.2 *
                       ---------
                          -973.2 Trade gain
                          -380.4 Best 22
                         +1124.7 vs last game
  Fringe player:
          Taylin Duman     134.5 (2021)
       Bailey Banfield      75.5 (2022)

  INS::::
         410.4 Nat Fyfe
         280.1 Michael Walters
         243.1 Rory Lobb
         230.4 Brennan Cox
         207.3 Nathan Wilson
         184.2 Lachie Schultz
         141.3 Ethan Hughes
     ------------------------------
         165.8 Mitchell Crowden (GONE)
         104.7 Brett Bewley (GONE)
          73.0 Hayden Young 
          66.9 Josh Treacy 
          61.0 Lloyd Meek 
          58.7 Tobe Watson (GONE)
          42.0 Brandon Walker 
  OUTS::::
       +1124.7 Net (Ins: +1696.8, Outs: +572.1)
That's pretty good -- currently ranked the 6th most off-season gain. You would expect some improvement off the back of that.

And Freo's inaccuracy last year is actually a good thing, in Squiggle's eyes, because it believes teams revert to the mean over time. (Even West Coast.) So if you're getting 30 shots on goal and kicking 10.20, Squiggle sees that as a team that's doing a lot right, and will probably get better results in the future as conversion normalizes.

But what's very bad is how Freo finished the year. Models put more faith in more recent results, and Fremantle after Round 17 were ordinary. Teams that know they're missing finals do tend to shut up shop late in the season, but they still often manage a few good results that suggest promise for next year. Those Round 18 - Round 23 performances are depressing predictions more than anything else, I'd say.
 
I think the model's aggregate ladder and what the media are saying about particular teams is quite different.

View attachment 1341341

I keep hearing in the media about how Carlton are going to play finals, and how Hawthorn are in danger of winning the wooden spoon. Yet, Hawthorn sit above Carlton on the aggregate ladder.
Squiggle is the least favorable one towards the Hawks in 2022 projecting a 16th place finish where every other algorithm have us higher than that.

35572FB3-A8CC-4AF0-9811-A3BA379DE444.png


At the end of 2021 Squiggle had Hawks 10th, and it seems doesn’t really rate the return of Gunston, Sicily, Impey or CJ, so the drop must be on the loss of Ceglar, Burgoyne and O’Brien(which is odd), as the coach isn’t a data point as far as I’m aware.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
I think the model's aggregate ladder and what the media are saying about particular teams is quite different.

View attachment 1341341

I keep hearing in the media about how Carlton are going to play finals, and how Hawthorn are in danger of winning the wooden spoon. Yet, Hawthorn sit above Carlton on the aggregate ladder.
Gotta say I side with the humans on this one. As per the above post, Squiggle's in-house model is the most pessimistic of all models on Hawthorn, tipping them for 16th, and as far as I know it's also the only one that tries to measure off-season list changes.

I have Hawthorn's off-season like this:
Code:
HAWTHORN
             Max Lynch     +14.7 
       Harrison Pepper      -0.0 
       Jonathon Patton      -0.0 
            Tom Scully      -0.0 
       Keegan Brooksby      -4.1 
         Damon Greaves     -40.9 
       Michael Hartley     -41.7 
       Oliver Hanrahan     -98.6 
         James Cousins    -140.2 *
        Shaun Burgoyne    -226.3 *
           Tim O'Brien    -236.2 *
       Jonathon Ceglar    -258.2 *
                       ---------
                         -1031.5 Trade gain
                          -398.8 Best 22
                          +284.7 vs last game
  Fringe player:
              Will Day     139.5 (2021)
       Lachlan Bramble      82.4 (2022)

  INS::::
         380.7 Luke Breust
         239.0 Dylan Moore
         177.8 Changkuoth Jiath
         153.8 Jack Gunston
         139.5 Will Day
         137.5 Jarman Impey
     ------------------------------
         258.2 Jonathon Ceglar (GONE)
         236.2 Tim O'Brien (GONE)
         226.3 Shaun Burgoyne (GONE)
          98.6 Oliver Hanrahan (GONE)
          69.6 Tyler Brockman 
          54.7 Conor Nash 
  OUTS::::
        +284.7 Net (Ins: +1228.3, Outs: +943.6)
... which is to say, their Round 1 team this year is expected to be better than their Round 23 team from last year, but not by much. In fact, because practically all teams improve over the off-season as they regain players from injury - often by a lot - a 284-pt gain is pretty anemic: the 3rd worst in the league.

Often the teams who have poor off-seasons were playing finals - they don't have much scope to add more talent for Round 1 because they were sending out their best teams late in the previous season, and didn't get draft picks to spend on recruitment. So the other bottom four teams are Melbourne, Port Adelaide, and Brisbane. It's a bit worrisome that after finishing 14th, Hawthorn have so little list improvement to show as they head into 2022.
 
Gotta say I side with the humans on this one. As per the above post, Squiggle's in-house model is the most pessimistic of all models on Hawthorn, tipping them for 16th, and as far as I know it's also the only one that tries to measure off-season list changes.

I have Hawthorn's off-season like this:
Code:
HAWTHORN
             Max Lynch     +14.7
       Harrison Pepper      -0.0
       Jonathon Patton      -0.0
            Tom Scully      -0.0
       Keegan Brooksby      -4.1
         Damon Greaves     -40.9
       Michael Hartley     -41.7
       Oliver Hanrahan     -98.6
         James Cousins    -140.2 *
        Shaun Burgoyne    -226.3 *
           Tim O'Brien    -236.2 *
       Jonathon Ceglar    -258.2 *
                       ---------
                         -1031.5 Trade gain
                          -398.8 Best 22
                          +284.7 vs last game
  Fringe player:
              Will Day     139.5 (2021)
       Lachlan Bramble      82.4 (2022)

  INS::::
         380.7 Luke Breust
         239.0 Dylan Moore
         177.8 Changkuoth Jiath
         153.8 Jack Gunston
         139.5 Will Day
         137.5 Jarman Impey
     ------------------------------
         258.2 Jonathon Ceglar (GONE)
         236.2 Tim O'Brien (GONE)
         226.3 Shaun Burgoyne (GONE)
          98.6 Oliver Hanrahan (GONE)
          69.6 Tyler Brockman
          54.7 Conor Nash
  OUTS::::
        +284.7 Net (Ins: +1228.3, Outs: +943.6)
... which is to say, their Round 1 team this year is expected to be better than their Round 23 team from last year, but not by much. In fact, because practically all teams improve over the off-season as they regain players from injury - often by a lot - a 284-pt gain is pretty anemic: the 3rd worst in the league.

Often the teams who have poor off-seasons were playing finals - they don't have much scope to add more talent for Round 1 because they were sending out their best teams late in the previous season, and didn't get draft picks to spend on recruitment. So the other bottom four teams are Melbourne, Port Adelaide, and Brisbane. It's a bit worrisome that after finishing 14th, Hawthorn have so little list improvement to show as they head into 2022.
This is what is very odd to most Hawk onlookers regarding the squiggle.

When Hawthorn knocked off Lions, Dogs and drew with the Dees they were missing the following players Gunston, Sicily, Day, Impey, Jiath, O’Meara(dogs), Worpel(dogs)… but somehow with those players back in the side, as well as further development, we will be no better off?

Makes no sense.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
This is what is very odd to most Hawk onlookers regarding the squiggle.

When Hawthorn knocked off Lions, Dogs and drew with the Dees they were missing the following players Gunston, Sicily, Day, Impey, Jiath, O’Meara(dogs), Worpel(dogs)… but somehow with those players back in the side, as well as further development, we will be no better off?

Makes no sense.
Sure, but equally you can ask why in the middle of those games, the Hawks also lost to Adelaide (in Melbourne), or got done by 10 goals by Freo in Tassie. Logically, it doesn't make much sense that you can beat a good team, then, with better players, lose to a worse team, yet it happens all the time. Ultimately there's a limit on how much you can learn from any single result. Models try to crunch all the (conflicting) data together and come out with a balance of probability.
 

Pessimistic

Cancelled
30k Posts 10k Posts HBF's Milk Crate - 70k Posts TheBrownDog
Sep 13, 2000
86,852
42,951
Melbourne cricket ground. Australia
AFL Club
Hawthorn
Other Teams
Horks
Squiggle is the least favorable one towards the Hawks in 2022 projecting a 16th place finish where every other algorithm have us higher than that.

View attachment 1341371

At the end of 2021 Squiggle had Hawks 10th, and it seems doesn’t really rate the return of Gunston, Sicily, Impey or CJ, so the drop must be on the loss of Ceglar, Burgoyne and O’Brien(which is odd), as the coach isn’t a data point as far as I’m aware.
Both models give the hawks 9 wins, 5% behind 12th only on percentage in squiggle. Squiggles forecasts do tend to over equalise in early projections
 
Aug 11, 2021
16
13
Melbourne
AFL Club
Geelong
Gotta say I side with the humans on this one. As per the above post, Squiggle's in-house model is the most pessimistic of all models on Hawthorn, tipping them for 16th, and as far as I know it's also the only one that tries to measure off-season list changes.

I have Hawthorn's off-season like this:
Code:
HAWTHORN
             Max Lynch     +14.7
       Harrison Pepper      -0.0
       Jonathon Patton      -0.0
            Tom Scully      -0.0
       Keegan Brooksby      -4.1
         Damon Greaves     -40.9
       Michael Hartley     -41.7
       Oliver Hanrahan     -98.6
         James Cousins    -140.2 *
        Shaun Burgoyne    -226.3 *
           Tim O'Brien    -236.2 *
       Jonathon Ceglar    -258.2 *
                       ---------
                         -1031.5 Trade gain
                          -398.8 Best 22
                          +284.7 vs last game
  Fringe player:
              Will Day     139.5 (2021)
       Lachlan Bramble      82.4 (2022)

  INS::::
         380.7 Luke Breust
         239.0 Dylan Moore
         177.8 Changkuoth Jiath
         153.8 Jack Gunston
         139.5 Will Day
         137.5 Jarman Impey
     ------------------------------
         258.2 Jonathon Ceglar (GONE)
         236.2 Tim O'Brien (GONE)
         226.3 Shaun Burgoyne (GONE)
          98.6 Oliver Hanrahan (GONE)
          69.6 Tyler Brockman
          54.7 Conor Nash
  OUTS::::
        +284.7 Net (Ins: +1228.3, Outs: +943.6)
How was Geelong's off-season out of personal interest.?
 
Aug 11, 2021
16
13
Melbourne
AFL Club
Geelong
Squiggle is the least favorable one towards the Hawks in 2022 projecting a 16th place finish where every other algorithm have us higher than that.

View attachment 1341371

At the end of 2021 Squiggle had Hawks 10th, and it seems doesn’t really rate the return of Gunston, Sicily, Impey or CJ, so the drop must be on the loss of Ceglar, Burgoyne and O’Brien(which is odd), as the coach isn’t a data point as far as I’m aware.
I have Hawthorn 11th and Brisbane 7th, something is clearly wrong.
 
Sure, but equally you can ask why in the middle of those games, the Hawks also lost to Adelaide (in Melbourne), or got done by 10 goals by Freo in Tassie. Logically, it doesn't make much sense that you can beat a good team, then, with better players, lose to a worse team, yet it happens all the time. Ultimately there's a limit on how much you can learn from any single result. Models try to crunch all the (conflicting) data together and come out with a balance of probability.
Correct, and this is why the squiggle is flawed as much as I love it.

Hawthorn got belted by the Dockers the same week that the transition was scuttled by the playing group. The squiggle cannot understand the impact of offfield factors in any way.

The other issue you raised highlights that the squiggle cannot recognise the way in which teams play, other than scoreboard output.

So for example, logic would suggest that if an injury riddled side can defeat 4 top 8 sides in the second half of the year, that the side wasn’t playing to its capability when it lost to all the bottom 6 sides that it did in the first half of the year, but the defensive nature of the plan and how it was flipped at the halfway mark isn’t really captured by the squiggle, unless the scoreboard ticks over.
 
Aug 11, 2021
16
13
Melbourne
AFL Club
Geelong
Correct, and this is why the squiggle is flawed as much as I love it.

Hawthorn got belted by the Dockers the same week that the transition was scuttled by the playing group. The squiggle cannot understand the impact of offfield factors in any way.

The other issue you raised highlights that the squiggle cannot recognise the way in which teams play, other than scoreboard output.

So for example, logic would suggest that if an injury riddled side can defeat 4 top 8 sides in the second half of the year, that the side wasn’t playing to its capability when it lost to all the bottom 6 sides that it did in the first half of the year, but the defensive nature of the plan and how it was flipped at the halfway mark isn’t really captured by the squiggle, unless the scoreboard ticks over.
At Round 5 the aggregate ladder is very good. I don't take it too seriously until then.
 

Trav 20

Brownlow Medallist
Oct 5, 2004
10,677
9,532
Idiots get ignored
AFL Club
Melbourne
Other Teams
Dees
Correct, and this is why the squiggle is flawed as much as I love it.

Hawthorn got belted by the Dockers the same week that the transition was scuttled by the playing group. The squiggle cannot understand the impact of offfield factors in any way.

The other issue you raised highlights that the squiggle cannot recognise the way in which teams play, other than scoreboard output.

So for example, logic would suggest that if an injury riddled side can defeat 4 top 8 sides in the second half of the year, that the side wasn’t playing to its capability when it lost to all the bottom 6 sides that it did in the first half of the year, but the defensive nature of the plan and how it was flipped at the halfway mark isn’t really captured by the squiggle, unless the scoreboard ticks over.
And you think the top sides played to their "capability" in the games they played Hawthorn ?

It's not uncommon for top sides to have mini lapses in a long season, especially when they come up against a lowly side. Their output isn't the same for every single round. Outliers happen.

Percentage across 22 games is the greatest indicator of a team's year and overall quality.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
Correct, and this is why the squiggle is flawed as much as I love it.

Hawthorn got belted by the Dockers the same week that the transition was scuttled by the playing group. The squiggle cannot understand the impact of offfield factors in any way.

The other issue you raised highlights that the squiggle cannot recognise the way in which teams play, other than scoreboard output.

So for example, logic would suggest that if an injury riddled side can defeat 4 top 8 sides in the second half of the year, that the side wasn’t playing to its capability when it lost to all the bottom 6 sides that it did in the first half of the year, but the defensive nature of the plan and how it was flipped at the halfway mark isn’t really captured by the squiggle, unless the scoreboard ticks over.
Squiggle does (attempt to) account for ins & outs, both week to week and over the off-season, but you're 100% right that there are a whole bunch of factors it doesn't account for. And a new coach is a big one!

A model is never going to be very good at predicting how well a team will go under a new gameplan. I'm not sure any of us are, either, but it's a situation that models are especially poor at, with very few historical data points to draw from, and lots of very important qualitative stuff.

Even the transition stuff is un-model-able. But I do get suspicious of narratives around things like that, because it's what humans love to do: invent stories to conjure meaning in a chaotic universe. There are plenty of teams that got thumped before/during/after a coaching transition, and plenty of teams that pulled off unexpected wins, and both ways, people rush to assign cause and effect. Even at the best of times, there's always a reason why this particular game was exceptional and shouldn't be taken as a true indicator of where the side is at.

On a personal note, I'm still emotionally scarred from when Jeff Gieschen took over from Robert Walls in 1997 and somehow went 5-0, securing himself the senior coaching gig before steering us into finals-free mediocrity for the next few years. New coaches, like new draft picks, bring hope and upside, but they crash and burn a lot, too.
 
Apr 23, 2016
30,510
42,671
AFL Club
Essendon
And you think the top sides played to their "capability" in the games they played Hawthorn ?

It's not uncommon for top sides to have mini lapses in a long season, especially when they come up against a lowly side. Their output isn't the same for every single round. Outliers happen.

Percentage across 22 games is the greatest indicator of a team's year and overall quality.

He'd be 100% behind Squiggle if it was pumping up Hawthorn.
 

Final Siren

Mr Squiggle
Aug 18, 2009
4,229
17,495
AFL Club
Richmond
I’d love to see Essendon’s pluses and minuses over the off season too!

Who did the best according to squiggle in the off season?
You're going to hate me, but Richmond:



So far -- can change with new injuries & I haven't done praccy match results yet -- off-season improvement looks roughly like this:

Tons: Richmond, Carlton
Plenty: West Coast, Fremantle, Collingwood, Gold Coast, Adelaide
Some: St Kilda, Essendon, Geelong, GWS, Bulldogs, Sydney, Nth Melb
Not much: Brisbane, Hawthorn, Port Adelaide, Melbourne
 
And you think the top sides played to their "capability" in the games they played Hawthorn ?

It's not uncommon for top sides to have mini lapses in a long season, especially when they come up against a lowly side. Their output isn't the same for every single round. Outliers happen.

Percentage across 22 games is the greatest indicator of a team's year and overall quality.
So the Swans, Giants, Dogs, Lions and Dees all just had off days. Great insight.
 
He'd be 100% behind Squiggle if it was pumping up Hawthorn.
Yes, because that’s exactly the point that AFLGlicko originally made.

The squiggle ranks Hawthorn lower than every other algorithm.
 

Trav 20

Brownlow Medallist
Oct 5, 2004
10,677
9,532
Idiots get ignored
AFL Club
Melbourne
Other Teams
Dees
So the Swans, Giants, Dogs, Lions and Dees all just had off days. Great insight.
So Hawthorn plays to their capability win they win and don't win they lose ?

Great insight.

Hawthorn were bottom 4 for percentage in season 2021, which is an accurate depiction of the sides quality and depth. Good luck in 2022. You'll need it.
 
Apr 23, 2016
30,510
42,671
AFL Club
Essendon
Yes, because that’s exactly the point that AFLGlicko originally made.

The squiggle ranks Hawthorn lower than every other algorithm.

Squiggle rates Hawthorn as 9.3 wins, as compared to the average of 9.5 wins. It's really not that different to the aggregate, and even the most optimistic model has 10.5 wins, or, 1.2 wins more than squiggle. It just rates other sides as more likely to improve beyond Hawthorn than the other models do, thus the lower ladder position.
 
This does nothing to explain Hawthorn losing to all those other sides.
Hawthorn were bottom 2 in the first half of the year, playing overly defensive football, letting sides start with a flurry.




At the halfway mark of the year the gameplan changed(for the third time in the last 5 years), Clarkson was ultimately moved on and our midfield set up was modified for the better with Mitchell’s influence(Newcombe and Nash).

4785C61F-A295-4E1B-BA9E-2613188F75E1.jpeg


👆 they’re not the stats of a side just catching some good sides off guard.

By the end of the year Hawthorn we’re not a bottom 2 side, by any stretch, but it remains to see what happens in 2022.
 
Back