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

[Richo83]

North, like most Docklands tenants, are hampered by the fact that when they play non-Victorian teams away, they travel to venues and states they only rarely visit, but when the roles are reversed, the interstate side is often quite familiar with Docklands (and travelling to Melbourne in general). Also, the Kangaroos spread their home games across two states, Victoria and Tasmania, which means they never build up the kind of overwhelming bank of familiarity that is enjoyed by most other teams.

Hang on, North are 16-6 at Hobart, they're 54-41 at docklands during the same period and 91-85 home and away. I don't see their Hobart games as a disadvantage given they often have much more experience playing at that ground than their opponents have. It's almost like playing at Kardinia for the cats. Yes, they don't get to play every game at docklands and don't develop continuity, but they, on average fair better at Tasmania than at any Victorian ground anyway. If I were just looking at trying to improve North's onfield performance and ignored financials and their ties to Victoria, I'd push for more Tasmanian games, not less. How many sides genuinely feel comfortable playing at a venue they've played maybe 2, 3, maybe 4 times at maximum (GWS only one) against a side which has played there 22 times? I know Richmond isn't. In fact, only nine teams have played in Hobart, meaning eight teams have zero experience at the venue compared to North's literal season equivalent of matches and eight years of experience at the venue. And given three of their four opponents at Hobart in 2020 have never played there, it's a massive legup. I'd be wiling to bet that they win at least three of their four next year.

 

[Final Siren]

So for the Tigers how does that translate?

Being a bugger (and hoping all works out in 2020), top 4 is teams near us, and everyone else is teams we should beat. Unfortunately it never works the way you expect.

Right - it's really not even worth calculating, because there's so much uncertainty. Whatever we determine now will certainly prove to be very wrong, as teams improve or decline next year.

You can do it in retrospect, though, after everyone's stopped caring! For example, this is what I have 2019:

​

Broken down by source of fixture advantage:

​

The types of advantage are:

1. Home Ground Advantage (HGA): Where you play.

2. Double-Up Games (Oppo): How strong the opponents are that you play twice.

3. Timing (Inouts): How strong your opponents are when you happen to play them (e.g. whether they have major Ins or Outs, or are going through a form slump / peak).

Home Ground Advantage was discussed a few posts earlier, and Double-Up Games everyone understands: it's tougher to play the Bulldogs twice than Gold Coast.

But timing is important, too, because in 2019, the Bulldogs were a more formidable opponent late in the season than early!

​

In 2019, Brisbane had the worst timing, catching a lot of teams at their peak, such as the Bulldogs and Richmond late. They played North in Round 2, before the wheels fell off, and then again in Round 18, just as the Roos were becoming good again.

Sydney and Richmond, by contrast, had excellent timing, mostly catching their opponents outside their peaks.

When you put it all together, GWS had the best 2019 fixture because they did well on all three measures: They had the 2nd-easiest double-up games (Gold Coast, Essendon, Hawthorn, Richmond, Sydney), the 5th best timing, and the 6th best Net Home Ground Advantage.

Sydney had a reasonably tough fixture if you only look at their double-up games (Geelong, GWS, Carlton, Essendon, Melbourne), but did well (in theory) on Home Ground Advantage and got exceptionally lucky on timing, so it was relatively good overall.

And North Melbourne had one of the hardest batches of double-up opponents in Brisbane, Essendon, Geelong, Hawthorn, and Port Adelaide - teams that finished 1st, 2nd, 8th, 9th, and 10th. That's a bit rough for a team that missed finals the year before and so was supposed to have a mid-tier draw! But that's what happens when a "bottom 6" team like Brisbane make the Top 4.

 

[Final Siren]

Hang on, North are 16-6 at Hobart, they're 54-41 at docklands during the same period and 91-85 home and away. I don't see their Hobart games as a disadvantage given they often have much more experience playing at that ground than their opponents have. It's almost like playing at Kardinia for the cats. Yes, they don't get to play every game at docklands and don't develop continuity, but they, on average fair better at Tasmania than at any Victorian ground anyway. If I were just looking at trying to improve North's onfield performance and ignored financials and their ties to Victoria, I'd push for more Tasmanian games, not less. How many sides genuinely feel comfortable playing at a venue they've played maybe 2, 3, maybe 4 times at maximum (GWS only one) against a side which has played there 22 times? I know Richmond isn't. In fact, only nine teams have played in Hobart, meaning eight teams have zero experience at the venue compared to North's literal season equivalent of matches and eight years of experience at the venue. And given three of their four opponents at Hobart in 2020 have never played there, it's a massive legup. I'd be wiling to bet that they win at least three of their four next year.

It's definitely better for North to drag a Melbourne-based team to Tassie. For example, in Round 23, 2020, the Kangaroos host the Demons at Bellerive. As they run out, it will be the Roos' 14th game in four years there and the Demons' third. So that's a large familiarity advantage.

That's the only time they play a Melbourne-based team there all year, though. Their other games are against non-Victorian teams. That's advantageous for North, of course, but it would have been advantageous at Docklands, too, by about the same amount. And under the model, at least, splitting those home games across two states means they don't accumulate really big familiarity numbers in either place - which I think is a pretty good approximation for what it does to the fan base and the crowd support at those matches.

North's Tassie home isn't rated the same as Geelong's Fortress Kardinia because the Cats play their ground more than twice as often. So even though in both cases we have a home team facing an opponent with very little ground experience, the model considers Geelong's situation to be worth a few more points because of that bulk of extra games. For the same reason, Port's 3 games in China aren't considered to be an overwhelming familiarity advantage over St Kilda's 1 - you need a lot of games to build up a loud and passionate home crowd.

I don't attempt to track individual teams' records at individual grounds, so you can definitely point to results and say that North have proven to be much stronger at Tassie than the model says. (Or Sydney are worse at home, for example.) It's fragile to do that, though, because those are small samples, and a few games can easily skew numbers one way or the other.

NORTH MELBOURNE (2020): -35.9

  +8.8  R23: Melbourne @ Bellerive Oval (TAS)
  +8.7  R19: Gold Coast @ Bellerive Oval (TAS)
  +7.8   R3: Port Adelaide @ Marvel Stadium (VIC)
  +6.4  R13: Brisbane Lions @ Bellerive Oval (TAS)
  +6.2  R16: Greater Western Sydney @ Marvel Stadium (VIC)
  +6.0   R5: Fremantle @ Bellerive Oval (TAS)
  -0.5   R8: Hawthorn @ Marvel Stadium (VIC)
  -0.9   R1: St Kilda @ Marvel Stadium (VIC)
  -1.0   R9: Essendon @ Marvel Stadium (VIC)
  -1.2   R4: Western Bulldogs @ Marvel Stadium (VIC)
  -1.2  R20: Carlton @ Marvel Stadium (VIC)
  -1.2  R21: Western Bulldogs @ Marvel Stadium (VIC)
  -1.6  R10: Collingwood @ Marvel Stadium (VIC)
  -2.5   R6: Carlton @ M.C.G. (VIC)
  -2.5  R18: Essendon @ M.C.G. (VIC)
  -3.2  R17: Hawthorn @ York Park (TAS)
  -4.1  R12: Richmond @ M.C.G. (VIC)
 -10.3  R11: Geelong @ Kardinia Park (Gee)
 -11.3   R2: Brisbane Lions @ Gabba (QLD)
 -12.1  R15: West Coast @ Perth Stadium (WA)
 -12.8  R22: Adelaide @ Adelaide Oval (SA)
 -13.4   R7: Sydney @ S.C.G. (NSW)

 

[Simon_Nesbit]

...<Excellent Anaylsis>

Hey FS, I'd love to see the figures for other teams - especially Hawthorn as I think it would help explain/disprove the Hawthorn = 2020 Finals posts, with a poor/atrocious early start and coming home strong to be one of the 4-5 best sides at years end.

(Or you can prove me wrong to my eternal embarrassment).

 

[Richo83]

It's definitely better for North to drag a Melbourne-based team to Tassie. For example, in Round 23, 2020, the Kangaroos host the Demons at Bellerive. As they run out, it will be the Roos' 14th game in four years there and the Demons' third. So that's a large familiarity advantage.

That's the only time they play a Melbourne-based team there all year, though. Their other games are against non-Victorian teams. That's advantageous for North, of course, but it would have been advantageous at Docklands, too, by about the same amount.

I'd be genuinely interested to see if Fremantle Brisbbane and Gold Coast, who play at docklands up to four times a year would rather play at a venue they're somewhat familiar with than a ground they've never seen with dimensions they've never experienced.

And under the model, at least, splitting those home games across two states means they don't accumulate really big familiarity numbers in either place - which I think is a pretty good approximation for what it does to the fan base and the crowd support at those matches.

I think you're assuming things about football teams, that is, that they need to play a lot of matches in a ground in a year to become familiar with it. I'm not so sure, and I still think that the experience they gain from playing there is vastly superior to the experience of other clubs.

Let me put it this way, North plays at docklands usually around 11 games a year. Those interstate sides play at docklands 2-4 times a year. North plays at Hobart 3-4 times a year, those three interstate sides play in Hobart maybe once in tree years or so? I'd take them longer to build up the level of experience at Hobart than at docklands. So why is docklands a benefit in a way that Hobart isn't? I mean if we're going to talk about developing familiarity with a ground, how is a side supposed to do that when they play there once every few years?

North's Tassie home isn't rated the same as Geelong's Fortress Kardinia because the Cats play their ground more than twice as often. So even though in both cases we have a home team facing an opponent with very little ground experience, the model considers Geelong's situation to be worth a few more points because of that bulk of extra games. For the same reason, Port's 3 games in China aren't considered to be an overwhelming familiarity advantage over St Kilda's 1 - you need a lot of games to build up a loud and passionate home crowd.

North's attendances, especially against interstate sides are bad and support sounds tend to dissipate when the attendance is low. At Hobart, they usually either sell out or have a fairly good crowd of North fans. I really doubt there's much difference in crowd support. Both crowds will be 100% North.

I don't attempt to track individual teams' records at individual grounds, so you can definitely point to results and say that North have proven to be much stronger at Tassie than the model says. (Or Sydney are worse at home, for example.) It's fragile to do that, though, because those are small samples, and a few games can easily skew numbers one way or the other.

Ok, but North have been playing at Hobart since 2012 and perform better than usual at the ground. It is data, and it does show that North play well there. There's also data to show that the level of experience vis a vis interstate sides at Hobart and docklands is roughly the same.

 

[Final Siren]

I think you're assuming things about football teams, that is, that they need to play a lot of matches in a ground in a year to become familiar with it.

Well, it's not really an assumption; the way I build algorithms is to brute force a ton of different values and see what works. The reason the model doesn't rate a 4:1 familiarity advantage quite the same way when it's 12 games to 3 compared to when it's 40 games to 10 is just because it seems to be more accurate that way.

That idea may be flawed, and in particular it might be picking up a third, unrelated variable! Like that teams who play 40 games in a single venue over 3 or 4 years mostly also just happen to have big, noisy supporter bases. In which case, North in Tassie in particular may consistently do better than the general model suggests.

Ok, but North have been playing at Hobart since 2012 and perform better than usual at the ground.

Before a model can believe that, it needs a general case, since it can't just think "North are better in Tassie." It needs to be something like, "Home Ground Advantage is stronger when the opposition has played very few games at the same venue." Then it's a rule that can be tested across all teams and venues, from a large data set.

So far, at least, I haven't seen better predictive power from assigning very high HGA to teams at grounds they don't play that often. Instead, it seems like the greatest HGA comes from grounds that the home team plays at 10 times a year or more. But HGA is a tricky thing and I could be wrong!

 

[Final Siren]

Hey FS, I'd love to see the figures for other teams - especially Hawthorn as I think it would help explain/disprove the Hawthorn = 2020 Finals posts, with a poor/atrocious early start and coming home strong to be one of the 4-5 best sides at years end.

(Or you can prove me wrong to my eternal embarrassment).

This one, you mean?

​

The Hawks' R1 defeat of Adelaide (in Adelaide) moved their rating a lot, since that can happen in the very early rounds. That aside, yeah, it's probably a general trend downward, i.e. becoming more dangerous late in the season, especially the last three rounds. You don't actually see where they finished on this chart because it doesn't have a post-R23 value, after they beat West Coast in Perth.

 

[tazaa]

Nah, he played for your mob.

Wish he still was

 

[Richo83]

Well, it's not really an assumption; the way I build algorithms is to brute force a ton of different values and see what works. The reason the model doesn't rate a 4:1 familiarity advantage quite the same way when it's 12 games to 3 compared to when it's 40 games to 10 is just because it seems to be more accurate that way.

That idea may be flawed, and in particular it might be picking up a third, unrelated variable! Like that teams who play 40 games in a single venue over 3 or 4 years mostly also just happen to have big, noisy supporter bases. In which case, North in Tassie in particular may consistently do better than the general model suggests.

How much data do we have analysing how many games teams need to play well at secondary grounds though?

Adelaide, Brisbane, Collingwood, Fremantle, Gold Coast, Port, Sydney, West Coast and Richmond all play at one location.

The exceptions are:

Carlton and Essendon, who split home matches in docklands and MCG. There's too much noise to figure out which ground is more relevant.

Geelong, who split games between Geelong, docklands and MCG. This is also the same team who complains incessantly whenever they have to play at home in Melbourne. To be fair, their record in Geelong is much better than Melbourne.

GWS, who split games between Sydney and Canberra. Seems like a fairly similar situation to North really.

Hawthorn, who split games between the MCG and Launceston. Their record at York Park is 49-1-14.

Footscay, who have played home games in Cairns, Darwin, Canberra and Darwin. Haven't really developed any continuity with any ground. They're startting to develop continuity with Ballarat, but they've just started.

Melbourne and St Kilda, who sometimes play in places like Darwin, Alice Springs, China, and NZ, but not regularly to count as a regular home ground.

And then there's North, with their regular matches at Melbourne and Tasmania.

In other words, there's only North, Hawthorn, Geelong and GWS who regularly play games at two specific occasions. Canberra is counted as a benefit for GWS, why not Tasmania for Hawthorn and North?

Before a model can believe that, it needs a general case, since it can't just think "North are better in Tassie." It needs to be something like, "Home Ground Advantage is stronger when the opposition has played very few games at the same venue." Then it's a rule that can be tested across all teams and venues, from a large data set.

So far, at least, I haven't seen better predictive power from assigning very high HGA to teams at grounds they don't play that often. Instead, it seems like the greatest HGA comes from grounds that the home team plays at 10 times a year or more. But HGA is a tricky thing and I could be wrong!

Again, that's because teams don't all have two home grounds (and in fact more clubs have exclusively one home ground than have two), meaning that making a reliable general rule to apply to all teams is a challenge given you're relying on new and inconsistent data (i.e. there's a difference between playing a game or two in a place here and there and playing three games consistently year in year out, even according to your own rules about familiarity).
I will say this though, it seems like GWS, Hawthorn and North have two genuine home grounds, and should be counted as such. Geelong don't have an advantage in Melbourne, and you count as much. Seems like a simple rule to apply?

 

[Final Siren]

How much data do we have analysing how many games teams need to play well at secondary grounds though?

I'm not sure I totally understand the question, but your post only seems to consider home teams. To apply a venue-and-team-specific model to West Coast v North Melbourne @ Perth Stadium, we need to know (1) how well West Coast play there, and (2) how well North play there. It's very hard to get any kind of reliable number for #2 because North play there so rarely, which means we have to choose between a tiny sample or else draw on old North Melbourne games that share very few of the same players (or staff).

And you have to look at both sides because it's a match between two teams. It would be clearly wrong to assign full HGA to the Eagles in a West Coast vs Fremantle game just because they're the home side, for example.

 

[flipper83]

Would love it if you could make a ladder predictor where we could adjust teams ratings. Say if we think Melbourne is a bit stronger than the squiggle rating we can adjust up and perhaps a team like Port or Hawthorn might be adjusted down slightly.
Not sure how difficult that would be to do?

 

[Final Siren]

Well Squiggle is definitely going to be bearish on the Eagles in 2020, relative to popular opinion.

Here's how many games each team won in 2019, plus the difference that was due to accurate/inaccurate goalkicking:



That is, if all teams converted at the league average of 52.06%, Melbourne would have won 9 games instead of 5, Hawthorn would have won 13 instead of 11, and West Coast would have won 10 instead of 15.

This is only the third time in the last 30 years that a team has recorded a difference of -5 or more. The two previous times were:

• Sydney 2018 with 14 wins (adjusted: 9). The following year, won 8 games, falling from 6th to 15th.
• Fitzroy 1993 with 10 wins (adjusted: 5). The following year, won 5 games, falling from 11th to 14th.

There have also been three occasions when the same magnitude of difference was recorded but in the opposite direction:

• West Coast 2014 with 11 wins (adjusted: 16). The following year, won 16 games, rising from 9th to 2nd.
• Richmond 2012 with 10.5 wins (adjusted: 15.5). The following year, won 15 games, rising from 12th to 5th.
• Fremantle 2008 with 6 wins (adjusted: 11). The following year, won 6 games again, remaining at 14th.

This is essentially showing that teams who win (or lose) a lot of games due to goalkicking conversion are unlikely to repeat it. Instead, they're more likely to perform at the kind of level they would have, had they converted more normally.

This thread already discussed goalkicking conversion in some depth this season, since Geelong and West Coast were both converting at an unusually high rate halfway through. Good teams do reliably convert better than bad ones - but aside from that, conversion isn't sustainable, and teams fall back to the pack sooner or later. (Usually sooner.) I'm not sure why - it could be luck that evens out, or opposition tactics adapting to counter a new weapon - but it inevitably doesn't last.

This caught up with Geelong in the second half 2019, as they fell back from 60% to finish at 55%. But West Coast managed to end the Home & Away season with a 57% conversion rate, which is unusually high even for a top team. Although, I should say, not outrageously so - in their previous four years, in all of which they made finals, the Eagles converted at 55%, 55%, 57% and 54%. Next year, they are likely to convert at a lower rate, but probably not a lot lower.

What makes West Coast 2019 really unusual is how often those high-converting games occurred just when they needed them. Even if the Eagles do record another 57% conversion rate in 2020, it would be remarkable for the same pattern to hold, where they tend to convert efficiently in games they would have otherwise lost, and convert less well in games where it wouldn't have made a difference to the result anyway.

​

This is a lot more about good timing than good conversion. Because if we replayed the season and granted West Coast a 57% conversion rate in every game while their opponents always converted at 52% - that is, we acknowledge that the Eagles can reliably convert a bit better than everyone else, but we don't give them any benefit from timing their especially high-converting games - they'd still have 5 fewer wins.

The Eagles would actually need to average 66% conversion - bearing in mind that 61% is the highest ever recorded - to win the same number of games they actually did in 2019 but without relying on good timing.

And it's worth noting that it wasn't just the Eagles kicking well: Their opposition also failed to kick well. This was similarly the case for Geelong. In fact, if you look at relative conversion - how well a team converted compared to their match-day opposition - the Cats and Eagles recorded two of the top three most favourable numbers since 2000.

St Kilda, incidentally, were hugely negative on the same metric in 2019, regularly converting much less efficiently than their opponents. This didn't have a huge impact on their season, though, since they only would have won one more game with league average conversion.

A lot of what a computer model does is try to weed out non-reproducible factors, and goalkicking conversion is a pretty big one. So although the Eagles are being talked up as a Top 4 lock and a likely flag favourite, Squiggle has them 9th.

 

[LittleG]

Well Squiggle is definitely going to be bearish on the Eagles in 2020, relative to popular opinion.

Here's how many games each team won in 2019, plus the difference that was due to accurate/inaccurate goalkicking:

View attachment 797429

That is, if all teams converted at the league average of 52.06%, Melbourne would have won 9 games instead of 5, Hawthorn would have won 13 instead of 11, and West Coast would have won 10 instead of 15.

This is only the third time in the last 30 years that a team has recorded a difference of -5 or more. The two previous times were:

• Sydney 2018 with 14 wins (adjusted: 9). The following year, won 8 games, falling from 6th to 15th.
• Fitzroy 1993 with 10 wins (adjusted: 5). The following year, won 5 games, falling from 11th to 14th.

There have also been three occasions when the same magnitude of difference was recorded but in the opposite direction:

• West Coast 2014 with 11 wins (adjusted: 16). The following year, won 16 games, rising from 9th to 2nd.
• Richmond 2012 with 10.5 wins (adjusted: 15.5). The following year, won 15 games, rising from 12th to 5th.
• Fremantle 2008 with 6 wins (adjusted: 11). The following year, won 6 games again, remaining at 14th.

This is essentially showing that teams who win (or lose) a lot of games due to goalkicking conversion are unlikely to repeat it. Instead, they're more likely to perform at the kind of level they would have, had they converted more normally.

This thread already discussed goalkicking conversion in some depth this season, since Geelong and West Coast were both converting at an unusually high rate halfway through. Good teams do reliably convert better than bad ones - but aside from that, conversion isn't sustainable, and teams fall back to the pack sooner or later. (Usually sooner.) I'm not sure why - it could be luck that evens out, or opposition tactics adapting to counter a new weapon - but it inevitably doesn't last.

This caught up with Geelong in the second half 2019, as they fell back from 60% to finish at 55%. But West Coast managed to end the Home & Away season with a 57% conversion rate, which is unusually high even for a top team. Although, I should say, not outrageously so - in their previous four years, in all of which they made finals, the Eagles converted at 55%, 55%, 57% and 54%. Next year, they are likely to convert at a lower rate, but probably not a lot lower.

What makes West Coast 2019 really unusual is how often those high-converting games occurred just when they needed them. Even if the Eagles do record another 57% conversion rate in 2020, it would be remarkable for the same pattern to hold, where they tend to convert efficiently in games they would have otherwise lost, and convert less well in games where it wouldn't have made a difference to the result anyway.

View attachment 797439​

This is a lot more about good timing than good conversion. Because if we replayed the season and granted West Coast a 57% conversion rate in every game while their opponents always converted at 52% - that is, we acknowledge that the Eagles can reliably convert a bit better than everyone else, but we don't give them any benefit from timing their especially high-converting games - they'd still have 5 fewer wins.

The Eagles would actually need to average 66% conversion - bearing in mind that 61% is the highest ever recorded - to win the same number of games they actually did in 2019 but without relying on good timing.

And it's worth noting that it wasn't just the Eagles kicking well: Their opposition also failed to kick well. This was similarly the case for Geelong. In fact, if you look at relative conversion - how well a team converted compared to their match-day opposition - the Cats and Eagles recorded two of the top three most favourable numbers since 2000.
View attachment 797438
St Kilda, incidentally, were hugely negative on the same metric in 2019, regularly converting much less efficiently than their opponents. This didn't have a huge impact on their season, though, since they only would have won one more game with league average conversion.

A lot of what a computer model does is try to weed out non-reproducible factors, and goalkicking conversion is a pretty big one. So although the Eagles are being talked up as a Top 4 lock and a likely flag favourite, Squiggle has them 9th.

So according to your graph, Dogs would have been up to second on the ladder behind Richmond IF goal kicking accuracy were equal for all teams.

 

[Sixpence]

So according to your graph, Dogs would have been up to second on the ladder behind Richmond IF goal kicking accuracy were equal for all teams.

Third, behind Tigers and Lions.

 

[threesixpio]

Well Squiggle is definitely going to be bearish on the Eagles in 2020, relative to popular opinion.

Here's how many games each team won in 2019, plus the difference that was due to accurate/inaccurate goalkicking:

View attachment 797429

That is, if all teams converted at the league average of 52.06%, Melbourne would have won 9 games instead of 5, Hawthorn would have won 13 instead of 11, and West Coast would have won 10 instead of 15.

This is only the third time in the last 30 years that a team has recorded a difference of -5 or more. The two previous times were:

• Sydney 2018 with 14 wins (adjusted: 9). The following year, won 8 games, falling from 6th to 15th.
• Fitzroy 1993 with 10 wins (adjusted: 5). The following year, won 5 games, falling from 11th to 14th.

There have also been three occasions when the same magnitude of difference was recorded but in the opposite direction:

• West Coast 2014 with 11 wins (adjusted: 16). The following year, won 16 games, rising from 9th to 2nd.
• Richmond 2012 with 10.5 wins (adjusted: 15.5). The following year, won 15 games, rising from 12th to 5th.
• Fremantle 2008 with 6 wins (adjusted: 11). The following year, won 6 games again, remaining at 14th.

This is essentially showing that teams who win (or lose) a lot of games due to goalkicking conversion are unlikely to repeat it. Instead, they're more likely to perform at the kind of level they would have, had they converted more normally.

This thread already discussed goalkicking conversion in some depth this season, since Geelong and West Coast were both converting at an unusually high rate halfway through. Good teams do reliably convert better than bad ones - but aside from that, conversion isn't sustainable, and teams fall back to the pack sooner or later. (Usually sooner.) I'm not sure why - it could be luck that evens out, or opposition tactics adapting to counter a new weapon - but it inevitably doesn't last.

This caught up with Geelong in the second half 2019, as they fell back from 60% to finish at 55%. But West Coast managed to end the Home & Away season with a 57% conversion rate, which is unusually high even for a top team. Although, I should say, not outrageously so - in their previous four years, in all of which they made finals, the Eagles converted at 55%, 55%, 57% and 54%. Next year, they are likely to convert at a lower rate, but probably not a lot lower.

What makes West Coast 2019 really unusual is how often those high-converting games occurred just when they needed them. Even if the Eagles do record another 57% conversion rate in 2020, it would be remarkable for the same pattern to hold, where they tend to convert efficiently in games they would have otherwise lost, and convert less well in games where it wouldn't have made a difference to the result anyway.

View attachment 797439​

This is a lot more about good timing than good conversion. Because if we replayed the season and granted West Coast a 57% conversion rate in every game while their opponents always converted at 52% - that is, we acknowledge that the Eagles can reliably convert a bit better than everyone else, but we don't give them any benefit from timing their especially high-converting games - they'd still have 5 fewer wins.

The Eagles would actually need to average 66% conversion - bearing in mind that 61% is the highest ever recorded - to win the same number of games they actually did in 2019 but without relying on good timing.

And it's worth noting that it wasn't just the Eagles kicking well: Their opposition also failed to kick well. This was similarly the case for Geelong. In fact, if you look at relative conversion - how well a team converted compared to their match-day opposition - the Cats and Eagles recorded two of the top three most favourable numbers since 2000.
View attachment 797438
St Kilda, incidentally, were hugely negative on the same metric in 2019, regularly converting much less efficiently than their opponents. This didn't have a huge impact on their season, though, since they only would have won one more game with league average conversion.

A lot of what a computer model does is try to weed out non-reproducible factors, and goalkicking conversion is a pretty big one. So although the Eagles are being talked up as a Top 4 lock and a likely flag favourite, Squiggle has them 9th.

WC underachieved last year. I totes expect them to fire up next year

 

[LittleG]

Third, behind Tigers and Lions.

Ok, third on percentage.

 

[Final Siren]

Ok, third on percentage.

Yep, with average conversion in 2019, the Dogs would have won 14.5 games instead of 12, finishing 3rd behind Richmond (15.5 wins) and Brisbane (15 wins).

The Dogs would have won these:

• R6: Freo 13.10 (88) def Bulldogs 9.15 (69)
• R14: Collingwood 13.4 (82) def Bulldogs 10.13 (73)

... and drawn these:

• R3: Gold Coast 10.13 (73) def Bulldogs 9.14 (68)
• R10: North Melbourne 18.7 (115) def Bulldogs 13.12 (90)

... and also drawn this instead of winning it:

• R15: Port Adelaide 5.11 (41) lost to Bulldogs 10.6 (66)

 

[richoatthedisco]

Squiggle is making the same mistake with the Eagles as those who say 'if only Brisbane kicked straight' in the first quarter of the qualifying final against Richmond. It assumes that a change in accuracy will have no effect on the number of scoring shots, yet we know that repeat entries create more opportunities. Brisbane/anti-Richmond fans lamented that if Brisbane had converted better they would have been far enough in front to perhaps change the result of the game. Yet as soon as Brisbane did finally score a goal, the ball went back to the centre, Richmond won the clearance and kicked a goal within 15 seconds. It is simplistic and plain wrong to project that better accuracy would turn 2.5 into 5.2, with nothing else changing.

It's the same in reverse with West Coast. Unless they are shown to be horrendous at locking the ball inside 50 and creating repeat chances, reduced accuracy would result in more scoring opportunities.

 

[blitzer]

Squiggle is making the same mistake with the Eagles as those who say 'if only Brisbane kicked straight' in the first quarter of the qualifying final against Richmond. It assumes that a change in accuracy will have no effect on the number of scoring shots, yet we know that repeat entries create more opportunities. Brisbane/anti-Richmond fans lamented that if Brisbane had converted better they would have been far enough in front to perhaps change the result of the game. Yet as soon as Brisbane did finally score a goal, the ball went back to the centre, Richmond won the clearance and kicked a goal within 15 seconds. It is simplistic and plain wrong to project that better accuracy would turn 2.5 into 5.2, with nothing else changing.

It's the same in reverse with West Coast. Unless they are shown to be horrendous at locking the ball inside 50 and creating repeat chances, reduced accuracy would result in more scoring opportunities.

I agree with this post so much. It continuously shits me when people claim their team would have won by miles if they kicked straight when the only reason they got so many shots on goals was because they were able to hold the ball in their 50 after kicking points. If your team is dominating in the centre AND also successfully forwarding pressing then maybe its a legitimate point but most times its a gross oversimplification. Not to mention often the reason teams are kicking lots of points is because they are only generating shallow entries which do not have the same value as getting a shot at goal from within 30 metres.

 

[Billy ray]

What would it take for suns to join the pack? Win their first 5 say by 30 each?

 

[Final Siren]

I agree with this post so much. It continuously shits me when people claim their team would have won by miles if they kicked straight when the only reason they got so many shots on goals was because they were able to hold the ball in their 50 after kicking points. If your team is dominating in the centre AND also successfully forwarding pressing then maybe its a legitimate point but most times its a gross oversimplification. Not to mention often the reason teams are kicking lots of points is because they are only generating shallow entries which do not have the same value as getting a shot at goal from within 30 metres.

This is absolutely true! But I don't think it invalidates the point I was trying to make before.

There are a couple of interesting areas here.

First, I don't mean to single out individual games and proclaim that one team "should" have won just because of conversion rate. It's always a little dangerous to talk about who "should" have won a single game. For starters, it depends what you mean by "should."

Take the 2008 Grand Final, where Hawthorn 18.7 (115) defeated Geelong 11.23 (89). We can argue who should have won in a variety of ways:

• The Hawks deserved to win, because they scored 26 more points
• But the Cats should have won, because they had 9 more scoring shots
• But the only reason the Cats had so many scoring shots was because the Hawks maniacally rushed behinds in a way that hadn't been done before, which blunted the Cats' attack
• But had they played again the next week, the Cats probably would have won, because they were a fundamentally stronger team (21-1 162%!) who would have adjusted to counter the tactic

Who "should" have won? If we're looking at this game in isolation, I think it's Hawthorn: They innovated, it worked, they won. But if we're trying to figure out which is the stronger team in a wider context, I think it's Geelong. And that's what a model is trying to detect: not who deserved the win, but how likely the team is to win again in the future.

Second, although you can point to individual games where one team deservedly won despite far fewer shots on goal - and you can definitely do that for individual quarters! - this argument becomes harder to sustain as you add more data. Because we can clearly see that goalkicking conversion rates reliably revert to a long-term average. Not because goalkicking is pure luck, but because when a team develops a method of reliably generating goals - whether it's a tactic like feeding runners out the back, or a single forward's skill, or whatever - their oppositions' priority must become to negate it.

So while a single game of oddly high goalkicking conversion might be due to tactics, gameplan, or some other kind of good play that deservedly won out, it's unlikely that a whole season of it will be reproduced. If it were, we should see teams sustaining seriously large gaps over the rest of the competition, at least occasionally, for years. Instead, we see all teams converting within a few percentage points of where their win rate would suggest, year after year, decade after decade - despite every team trying its hardest to maximize that measure, over thousands of games and tens of thousands of scoring shots.

Therefore I agree that you can't point to the scoring shots of a single game as proof of much, but still, I feel pretty confident that when a team records season-long numbers that are wildly out of step with the long-term average, it's unlikely to do it again the following year.

Also! Just incidentally, obviously this isn't a binary choice between:
(1) Points scored is the best measure of true team performance
(2) Scoring shots is the best measure of true team performance

The truth is in the middle. Conversion rate is tied to quality of shots, and strong teams reliably generate higher quality chances while preventing their opposition from doing the same. This is why Champion Data (and other models) use "Expected Score," which is neither (1) nor (2) but a kind of hybrid, and considers what a typical team would score on average from those shots, taking into consideration distance, angle, type of kick, and pressure. The idea is to strip out the non-reproducible part (short-term conversion rate), leaving the reproducible part (scoring chances generated).

Squiggle doesn't have access to CD's shot quality data, but it tries to proxy the same thing by awarding behinds about 2.5pts and goals about 4.5pts. This too is a midpoint between (1), where goals are 6 times better than behinds, and (2), where goals and behinds are treated as identical.

 

[WillHayward]

So we should be getting on the Dogs, Dees and Hawks in 2020?

 

[Pessimistic]

This is absolutely true! But I don't think it invalidates the point I was trying to make before.

There are a couple of interesting areas here.

First, I don't mean to single out individual games and proclaim that one team "should" have won just because of conversion rate. It's always a little dangerous to talk about who "should" have won a single game. For starters, it depends what you mean by "should."

Take the 2008 Grand Final, where Hawthorn 18.7 (115) defeated Geelong 11.23 (89). We can argue who should have won in a variety of ways:

• The Hawks deserved to win, because they scored 26 more points
• But the Cats should have won, because they had 9 more scoring shots
• But the only reason the Cats had so many scoring shots was because the Hawks maniacally rushed behinds in a way that hadn't been done before, which blunted the Cats' attack
• But had they played again the next week, the Cats probably would have won, because they were a fundamentally stronger team (21-1 162%!) who would have adjusted to counter the tactic

Who "should" have won? If we're looking at this game in isolation, I think it's Hawthorn: They innovated, it worked, they won. But if we're trying to figure out which is the stronger team in a wider context, I think it's Geelong. And that's what a model is trying to detect: not who deserved the win, but how likely the team is to win again in the future.

Second, although you can point to individual games where one team deservedly won despite far fewer shots on goal - and you can definitely do that for individual quarters! - this argument becomes harder to sustain as you add more data. Because we can clearly see that goalkicking conversion rates reliably revert to a long-term average. Not because goalkicking is pure luck, but because when a team develops a method of reliably generating goals - whether it's a tactic like feeding runners out the back, or a single forward's skill, or whatever - their oppositions' priority must become to negate it.

So while a single game of oddly high goalkicking conversion might be due to tactics, gameplan, or some other kind of good play that deservedly won out, it's unlikely that a whole season of it will be reproduced. If it were, we should see teams sustaining seriously large gaps over the rest of the competition, at least occasionally, for years. Instead, we see all teams converting within a few percentage points of where their win rate would suggest, year after year, decade after decade - despite every team trying its hardest to maximize that measure, over thousands of games and tens of thousands of scoring shots.

Therefore I agree that you can't point to the scoring shots of a single game as proof of much, but still, I feel pretty confident that when a team records season-long numbers that are wildly out of step with the long-term average, it's unlikely to do it again the following year.

Also! Just incidentally, obviously this isn't a binary choice between:
(1) Points scored is the best measure of true team performance
(2) Scoring shots is the best measure of true team performance

The truth is in the middle. Conversion rate is tied to quality of shots, and strong teams reliably generate higher quality chances while preventing their opposition from doing the same. This is why Champion Data (and other models) use "Expected Score," which is neither (1) nor (2) but a kind of hybrid, and considers what a typical team would score on average from those shots, taking into consideration distance, angle, type of kick, and pressure. The idea is to strip out the non-reproducible part (short-term conversion rate), leaving the reproducible part (scoring chances generated).

Squiggle doesn't have access to CD's shot quality data, but it tries to proxy the same thing by awarding behinds about 2.5pts and goals about 4.5pts. This too is a midpoint between (1), where goals are 6 times better than behinds, and (2), where goals and behinds are treated as identical.

geelong 2008 supposedly one of the best sides ever but mooney and lonergan as kpp? Chappy and stokes below par and johnson nowhere to be seen?

its a historically unimpressive fwd line, surely the midfield kicked the goals in the regular season. clarko had a plan to restrict them with superior running power. Young changa crawford.
maybe the cats main game had been so successful they hadnt got a plan B

though young, the hawthorn forward line was far superior, and the defence didnt need to be

also I reckon the cats won just nine quarters in their 4 grand finals. A bit lucky to have 3 flags just quietly

 

[Final Siren]

How much do preseason matches matter?

Just a quick observation of the 2019 pre-season... with only two games played per team, people were even quicker than usual to dismiss the whole thing as a bit of kick-and-giggle with no predictive value. But was it?

Here is every team ranked from best to worst based on how they under- or over-performed vs expectation using the Squiggle model in their two preseason games:

Rank	Team	2019 Preseason	2019 H&A Results
1.	Carlton	def. Adelaide; lost to Collingwood by 4pts	Rose from 18th to 16th, winning 5 more games.
2.	Brisbane	thrashed Hawthorn; def. Melbourne	The surprise packet of the year, rising from 15th to 2nd, winning 11 more games.
3.	Adelaide	def. Port; def. GWS	Trod water, rising one spot from 11th to 10th, but winning 2 fewer games.
4.	Gold Coast	def Bulldogs; lost to Sydney	Started the year on fire, winning 3 of their first 4 games and losing the other by a point, then crashed, sliding one run to 18th and winning 1 game fewer.
5.	St Kilda	def North; def Bulldogs	Rose from 16th to 14th, winning 4.5 more games.
6.	Richmond	def Melbourne; def Hawthorn	Slid from 1st to 3rd, winning 2 fewer games, but things worked out okay.
7.	Geelong	lost to West Coast; def Essendon	Rose from 8th to 1st, winning 3 more games.
8.	Fremantle	lost to Collingwood; lost to West Coast	Rose from 14th to 13th, winning 1 more game.
9.	West Coast	def Geelong; def Fremantle	Slid from 2nd to 5th, winning 1 fewer game.
10.	Sydney	lost to GWS; thrashed Gold Coast	Crashed from 6th to 15th, winning 6 fewer games.
11.	Port Adelaide	def North; def Adelaide	Remained 10th, winning 1 fewer game.
12.	Western Bulldogs	lost to Gold Coast; lost to St Kilda	Overcame a sluggish first half of the season to rise from 13th to 7th, winning 4 more games.
13.	North Melbourne	lost to St Kilda; lost to Port	Slid from 9th to 12th, winning 2 fewer games.
14.	Hawthorn	thrashed by Brisbane; lost to Richmond	Slid from 4th to 9th, winning 4 fewer games.
15.	GWS	def. Sydney; lost to Adelaide	Rose from 7th to 6th despite winning 0.5 fewer games.
16.	Collingwood	def. Fremantle; barely beat Carlton	Slid from 3rd to 4th, winning the same number of games.
17.	Essendon	lost to Carlton; lost to Geelong	Rose from 11th to 8th despite winning the same number of games with a 10-pt inferior percentage.
18.	Melbourne	lost to Richmond; lost to Brisbane	Crashed from 5th to 17th, winning 9 fewer games.

Obviously this is far from perfect, but as a rough guide to who's going to have a better year and who's going to have a worse one, it's not too shabby!

Squiggle has factored in pre-season results for a couple of years now, because they really do seem to contain some useful signal about how well prepared a team is for the real thing.

 

[boondy]

FYI this is how Squiggle sees the 2020 fixture in terms of Home Ground Advantage.

No consideration to strength of double-up opponents here -- just advantage derived from who plays where.

View attachment 781340​

Breaking it down by individual games:

View attachment 781346

Note:

This is based on Squiggle's Ground Familiarity Model. See further down for details. It naturally tends to assign greater HGA to non-Victorian teams because they are more familiar with their opponents' home grounds than their opponents are with theirs. You may or may not agree with this methodology.

Factual observations:

Every team plays at least one game in Victoria and Western Australia.

Collingwood are the only team to not play in South Australia.

Melbourne and West Coast don't visit NSW.

The most games in a single state is 17, with Collingwood, Carlton, and the Bulldogs in Victoria. In the Bulldogs' case, one of these is in Geelong and two more are in Ballarat; Collingwood and Carlton only play in Melbourne. Two teams have 16 games in Victoria (all in Melbourne): Richmond and Essendon.

The least number of games played in a home state is Gold Coast with 11 in Queensland. Unless you want to treat Geelong as its own state, in which case it's Geelong with 9. All other teams have at least 12 games in their home state.

The most number of games at a single venue is 14, shared by three teams: Richmond (MCG), Collingwood (MCG), and St Kilda (Docklands). Melbourne have 13 MCG games. Five teams have 12 games at a single venue: West Coast (Perth), Adelaide (AO), the Bulldogs (Docklands), Port Adelaide (AO), and Fremantle (Perth).

The least number of games at a primary home venue is 8: GWS at Sydney Showgrounds and Hawthorn at the MCG. Three teams have only 9 games at their primary venue: Geelong (Kardinia), Essendon (Docklands), and North Melbourne (Docklands).

The team playing at the most number of different venues (by far) is Gold Coast with 12. Most teams play at 8 or 9 different venues. Two teams play at only 7 venues: Richmond and Essendon. Two teams play at only 6 different venues: Collingwood and Carlton.

The most number of different areas (treating China, Geelong, and the NT as separate areas to the states) is 8, shared by three teams: Gold Coast, Adelaide, and St Kilda. Gold Coast and Adelaide are the only teams to play in all 6 states plus the NT and Geelong. The least number of different areas is 4, with Collingwood only playing in Melbourne, Perth, Sydney, and Brisbane.

Model observations:

North, like most Docklands tenants, are hampered by the fact that when they play non-Victorian teams away, they travel to venues and states they only rarely visit, but when the roles are reversed, the interstate side is often quite familiar with Docklands (and travelling to Melbourne in general). Also, the Kangaroos spread their home games across two states, Victoria and Tasmania, which means they never build up the kind of overwhelming bank of familiarity that is enjoyed by most other teams.

Melbourne give away a stack of HGA by hosting Adelaide in the Northern Territory instead of at the MCG, and are additionally dragged to Tasmania by North in what would otherwise be a game at a fairly neutral venue.

Similarly, Gold Coast give away home advantage by hosting St Kilda in the NT.

In all of Geelong's games, they are either at some kind of disadvantage (MCG, Docklands, non-Vic) or else have a very large advantage (Kardinia).

Adelaide dodge a game of major disadvantage by playing Melbourne in the NT.

Similarly, Port Adelaide play away to St Kilda in China, which is a relatively neutral venue for what would other be a major disadvantage.

West Coast, like all non-Victorian teams, enjoy a mild benefit from playing multiple away games at somewhat familiar Victorian venues. Additionally they are hosted by Carlton at the MCG, which isn't Carlton's home ground.

GWS only have to travel to play non-Victorian sides twice. That is, they have 20 games in NSW and Victoria. This is a very good thing, under a ground familiarity model, since it allows them to build up familiarity with the same venues.

Ground Familiarity Model

There are a few different well-respected methods of calculating Home Ground Advantage; this is Squiggle's Ground Familiarity model, which scores teams based on how frequently they've played at the same ground and the same state relative to each other in recent years, plus a small modifier for interstate travel. It's a conservative and reliable method that treats all teams the same, and doesn't attempt to measure how well individual teams play individual grounds (which is fraught with danger).

In reality, the major factor in Home Ground Advantage seems to be crowd noise (which may influence umpiring and player psychology). This is hard to measure directly, so a model such as Ground Familiarity acts as a proxy for this - teams tend to have larger, noisier supporters when they're playing at grounds they play a lot.

It's not perfect by any measure, though, and like any assessment of Home Ground Advantage, should be best treated as a general guide - accurate to within a goal or so, but probably not much more than that.

So to sum it all up a GF in victoria is a 10 goal advantage to vic teams .