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

[Yippy Yi Yeo]

Squiggle did actually become player-aware halfway through this year, i.e. it now does account for injuries.

But teams' chart positions don't describe how strong they would have been with no injuries; they describe how strong the team actually was. So in that sense, the Eagles are indeed rated lower than they would have been with no injuries. (Same as with any team.)

Squiggle hasn't made a 2019 prediction yet because the fixture came out while I was on holidays, but once I've had a chance to run it through, it will account for personnel changes, including returns from injury.

Interesting, I wasn't aware of that as I didn't closely follow the squiggle this year because I was overseas. How much more accurate was it when incorporating the player ratings compared to without?

 

[RookiePick]

Squiggle did actually become player-aware halfway through this year...

Am I the only person who envisaged robot tanks driving atop piles of human skulls whilst reading this post?

 

[Dr Tigris]

The squiggle is a measure of probability and when the 40% outcome happens the squiggle is 'wrong' or 'underrating a team'? No. Weaker teams win sometimes, it's called chance. If chance didn't exist then why bother to play at all?

Probabilistic thinking screws many/most peoples' brain up. e.g. 538 gave █████ a 30% chance of winning the US election. He did whilst losing on number of votes. So that he won isn't against the prediction. It's just that he was an outside shot. And outside shots win quite often.

That the Tigers were favourites this year and played a really poor game against a red hot Pies can't be accurately predicted. But it wasn't from another dimension. We all knew the Pies were classy. We all knew the Tigers could play poorly. Equals 'upset'. Don't know the chance's pre-game off hand, but not unbelievable.

WC was a team that didn't really set the world on fire in 2018, but they just win the games that mattered. It could be pure luck - Sheed's kick was skill, but maybe luck overall. Doesn't matter. WC won the title. Good on them. Probabilistically there is a chance that a team like WC will win every year. Usually it is favourites win. But not always. That's 'because' of the probabilistic nature of footy and reality.

 

[Final Siren]

Interesting, I wasn't aware of that as I didn't closely follow the squiggle this year because I was overseas. How much more accurate was it when incorporating the player ratings compared to without?

Lots, but mostly because of luck!

The SQUIGGLE2 algorithm scored 142 tips at 68.6% for the year, while the new player-aware SQUIGGLE4 recorded 151 tips at 73.0%. So things improved markedly when I switched over mid-season.

This is a big gap -- 142 was poor for 2018, while 151 was exceptional -- but raw numbers of correct tips isn't a very reliable measurement, as a small number of coin-toss games can have an outlandishly large influence on how well you do. And that's what happened: SQUIGGLE2 got a bit unlucky with close games while SQUIGGLE4 got a bit lucky. This is evidenced by how they performed on more reliable measurements like Mean Average Error and Bits: on these, SQUIGGLE4 was better, but not by much.

I'm confident that SQUIGGLE4 is a superior algorithm, but it will only net about 2 more correct tips per year on average, and some years it will do worse.

 

[Unique Name]

Lots, but mostly because of luck!

The SQUIGGLE2 algorithm scored 142 tips at 68.6% for the year, while the new player-aware SQUIGGLE4 recorded 151 tips at 73.0%. So things improved markedly when I switched over mid-season.

This is a big gap -- 142 was poor for 2018, while 151 was exceptional -- but raw numbers of correct tips isn't a very reliable measurement, as a small number of coin-toss games can have an outlandishly large influence on how well you do. And that's what happened: SQUIGGLE2 got a bit unlucky with close games while SQUIGGLE4 got a bit lucky. This is evidenced by how they performed on more reliable measurements like Mean Average Error and Bits: on these, SQUIGGLE4 was better, but not by much.

I'm confident that SQUIGGLE4 is a superior algorithm, but it will only net about 2 more correct tips per year on average, and some years it will do worse.

Have you thought or tinkered with other additional inputs? Anything in the works?

 

[Forward Press]

Looking forward to Squiggle 4, also was not aware that you incorporated AFL Player Ratings into it halfway through last season.

I wonder what else can be included without compromising the model too much. Historical head to head results?

 

[Final Siren]

Have you thought or tinkered with other additional inputs? Anything in the works?

I'm looking at tracking whether teams lose players during a match. If you only get 50% Time On Ground out of a key player or two, that will hurt your performance, but the Squiggle model is unaware of it. Squiggle can even get further off-track if it sees the injured players are out of the team the following week, because it anticipates even worse performance -- when in fact it may be better, because the injured players are being replaced by guys who can actually take to the field.

This may go nowhere, though; a lot of these investigations wind up with me finding nothing very significant.

 

[Isaac Cumming No 1]

I'm looking at tracking whether teams lose players during a match. If you only get 50% Time On Ground out of a key player or two, that will hurt your performance, but the Squiggle model is unaware of it. Squiggle can even get further off-track if it sees the injured players are out of the team the following week, because it anticipates even worse performance -- when in fact it may be better, because the injured players are being replaced by guys who can actually take to the field.

This may go nowhere, though; a lot of these investigations wind up with me finding nothing very significant.

Would have boosted us on the flagpole last year no doubt. Finishing a game with a fit bench was a rarity.

I wonder if the prediction of a worse performance the following week isn't somewhat accurate. The players do get replaced by fit ones. We had a few games like round 22 against the Swans where the players just looked like they had tired legs from the extra strain though.

Just a thought. I'm certainly no expert on those things.

 

[Final Siren]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

 

[Billy ray]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

5 wins I’ll take!!

 

[Cuzz09]

Equal top [emoji108] Need to put teams away and build that %.

Seems like having the foot ln the throat attitude will be important next year

 

[jarrod_island_doggies]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

So *this* is the year the Swans fall off the cliff, huh? Bold strategy to predict a 15th place finish, let's see if it pays off (inb4 they finish top 6 for the rest of our natural lives...)

 

[the harry]

I'm looking at tracking whether teams lose players during a match. If you only get 50% Time On Ground out of a key player or two, that will hurt your performance, but the Squiggle model is unaware of it. Squiggle can even get further off-track if it sees the injured players are out of the team the following week, because it anticipates even worse performance -- when in fact it may be better, because the injured players are being replaced by guys who can actually take to the field.

This may go nowhere, though; a lot of these investigations wind up with me finding nothing very significant.

There was an article a whole back about one of the first guys to apply algorithms to racing in Asia. One part detailed how he was site weather would have been an impacting factor and so he travelled to England and copied, by hand, historical weather information only to find out had no real statistical significance.

 

[Dr Tigris]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

Very even looking year.

With the Swans falling off a cliff. Hmm. That prediction has been made many times.

 

[Forward Press]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

Would love it if there was just two games between finishing as minor premiers and finishing with a top 10 draft pick.

 

[Proper Gander]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

I'll take it. I was actually really looking forward to 2019 season but I would cheerfully agree to ditching the H&A and just substituting the squiggle ladder. I'll be much less tense and read more novels.

Melbourne first and Hawks missing the 8 - I can live with it as it is. I'm starting to already feel sick from squiggle induced pressure though.

 

[TheBigLeftBoot]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

 

[Joao]

We are doomed to win 8 games for the year yet again. Get pick 3 for it this time though (and with a better %).

 

[Final Siren]

So *this* is the year the Swans fall off the cliff, huh? Bold strategy to predict a 15th place finish, let's see if it pays off (inb4 they finish top 6 for the rest of our natural lives...)

There's never been a better time to be bearish on the Swans!

1. 2018 Overperformance

Sydney finished 6th in the regular season, but had the worst percentage of any finalist. They were five percentage points behind than the second-worst team (GWS) and ten percentage points behind the third-worst (Hawthorn). Which is a lot.

Like Fremantle in 2015, they started the year well enough but tailed off badly. Their finishing position was flattered by a collection of close wins -- they won only 4 games after the bye, and these were by 6, 2, 9, and 20 -- and unusual goalshooting accuracy, including kicking 11.1 against Richmond and beating both Melbourne and Collingwood despite having significantly fewer scoring shots. They were also trounced by Essendon and Gold Coast, and then obliterated by GWS (the 6th strongest finalist) in their first and only final, kicking just four goals for the match.

The Swans did make finals comfortably, finishing 2 wins and a smidge of percentage ahead of 9th. But this was almost entirely due to their early-season form, which had them 2nd on the ladder after Round 14. After that, the wheels well and truly fell off, to the point where they finished rated as only the 12th best team.

This is all quite reminiscent of Fremantle 2015, who won the minor premiership but were rated only 10th by Squiggle, and then finished bottom-4 when reality caught up to them the next year.

2. Lack of Injury Upside

Every team will welcome back important players next year who weren't able to get on the park by the end of 2018. In most cases, this should make a real difference, because they were missing a lot of players, or a couple of very good ones. But three clubs were able to name sides late in 2018 that are basically identical to their best 22 today: North Melbourne, Sydney, and Richmond.

To be fair, the Swans had a couple of key players running around in the final who were clearly hampered by injury. (Likewise Richmond.) But still, it means the club was able to put out something like its best complement of soldiers even late in the season, and so was operating near its peak.

3. Failure to Improve Best 22

This is actually looking better for Sydney than when I posted yesterday, because now they've announced they're picking up Daniel Menzel. Even so, they're currently up 3 players, having gained Menzel, Ryan Clarke, Jackson Thurlow, but down 9 (Dan Hannebery, Nic Newman, Gary Rohan, Dean Towers, Harry Marsh, Dan Robinson, Kurt Tippett, Jordan Foote, Alex Johnson).

Most of those losses don't matter, because they don't affect the best 22. Collingwood and Richmond, for example, both improved their position because they jettisoned a lot of players who aren't in the best 22 and picked up one or two who are. That won't make a difference unless their depth is tested.

Looking only at best 22, the Swans are up 2 (Menzel and Clarke) and down 4 (Hannebery, Newman, Rohan, Towers). This means two fringe players have just been promoted into Sydney's best squad.

So...

Sydney currently exhibit less 2019 upside than every other team but Gold Coast. That's not flattering: The Suns ran a chainsaw through their list this off-season, effectively losing half of their best 22. The only thing that gives Gold Coast any upside at all is that they were badly weakened by injury in late 2018, and will regain those players next year.

Sydney don't have injured players to bring back. They're rated 12th on form but have little apparent room for improvement. So they are expected to be overtaken by teams around the same area with more upside, such as the Bulldogs, Brisbane, and St Kilda.

 

[cryptor]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

If Swans do finish 15th then it will be the lowest placing on the ladder that Buddy Franklin has seen the team he plays for finish during his AFL career. But if you want to talk relative to the bottom of the ladder (given 2 new teams have joined since he was drafted) then that would be the lowest (4th last) since his debut year when Hawthorn finished 14th (3rd last).

A pretty poor place for him to see them finish given he chose to go to them for what he saw was a club he would win more flags with. Instead he'll have added no premiership medallions to his otherwise long list of honours and will be faced with the prospect of another 3 seasons with a club who would then realistically be facing the prospect of a rebuild. But at least he'll be getting paid well.

Personally, I don't see the Swans finishing anywhere near that low. Yeah, they lost a few best 22 players but I think many of those that are still there underperformed in 2018. I would expect those players to bounce back and the Swans to finish at least above 12th place.

 

[Turtles all the way down]

Am I the only person who envisaged robot tanks driving atop piles of human skulls whilst reading this post?

dundun dun dundun

 

[Forward Press]

It's Sydney, they haven't finished in the bottom 4 in god knows how long.

I am always wary of writing them off.

 

[Cuzz09]

Yeah Sydney and Adelaide just very rarely totally bottom out.

 

[Lsta062]

Very even looking year.

With the Swans falling off a cliff. Hmm. That prediction has been made many times.

Sydney are a bit like West Coast in that they rise whenever the consensus among the AFL community becomes that they’ll fall. The difference here is just that they rarely fall out of the 8 nowadays.

 

[Handsome.B.Wonderful]

Still a work in progress, but this is how the 2019 prediction is shaking up, factoring in trades & players returning from injury.

The number of predicted wins is smooshed in toward the middle -- in reality, some teams will certainly win more than 14 games -- since we can't really know who will break away from the pack. The number in the far-right column is the best indicator of what Squiggle thinks.

Is it just me or does this total 199 wins?