Certified Legendary Thread Race for the flag, in squiggly lines
- Thread starter Final Siren
- Start date
- Tagged users None
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,227
People are right to pay attention to prelim form; it's usually a very good indicator.
The Monday's experts opinion seems to be that everyone got carried away with Sydney's prelim win over North Melbourne and failed to realize that the Roos just weren't that good.
But that's too simplistic; many people (most?) thought North were a relatively weak prelim finalist, but that doesn't mean they were witches' hats. There's a lot of middle ground between "beating North means Sydney will win the flag" and "beating North doesn't mean anything."
Also of course Sydney looked a top side even before the prelims.
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,228
I had a little poke around, just for you. Unfortunately the lower scores in NFL do cause some wild distortions.
When an NFL match delivers a scoreline like 6-34, the squiggle sees one side scoring 5.7 times more than the other. This would be a monumental coach-killing landslide in AFL, like 42-272, but it doesn't seem that unusual for NFL. Teams regularly double their opponents' score, and an extra field goal here or there can change a good prediction into one that's out by 20% or more.
It also means that squiggle quirk where holding a team to a very low score can cause a very large jump in their position becomes a bigger issue. Very low scores are a kind of twilight zone for the squiggle, since it sees keeping a team to 40 points as twice as good an effort as keeping them to 80, and keeping them to 20 points is 4 times as good, and keeping them to 10 is 8 times as good, and keeping them to 5 points is 16 times as good... at which point things are getting pretty ridiculous, but nobody gets kept to 5 or 10 points any more, so it doesn't matter. But in NFL teams do get kept to 3 or 6 points, and even 0, which would cause the squiggle to believe its opponent's defence is infinitely good.
So essentially I don't think the basic squiggle model, which relies on percentages ("Richmond scored 16% more than predicted and held their opponents to 30% less"), would hold up well.
SJ
Premium Platinum
What about differentials? Or a combination of both?
Perhaps this would work better for footy too?
tribey
ʎǝlʞuᴉH ʞɔɐS
Veteran
Port Adelaide - Connor Rozee Player Sponsor 2020
Port Adelaide - Zak Butters 2019 Player Sponsor
Port Adelaide - Jack Watts Player Sponsor 2018
Port Adelaide - Nathan Krakouer Player Sponsor 2015
Port Adelaide - Matthew Broadbent Player Sponsor 2013
Port Adelaide - John Butcher 2012 Player Sponsor
Were also staring down the barrel of going 2-5 after trailing all day at home to Essendon in Round 7. Fought back to win by a kick after being 16 points down at 3QT.
Were soundly beaten by Adelaide (41pts, home), Melbourne (34pts, home) & WCE (45pts, away) in the previous three weeks. Bizarre.
Including missed free kick for Keneally holding Ashley sampi while Leo Barry takes that mark.
Yeah by almost 2 goals when Ablett passed it to Booze Bus Benny and the Hunter scored minutes later. He put it through to go in front. We were staring down the barrel at that point, and it was almost identical to 2012 when we went 2 goals down almost straight away in the last.
Hall got a goal, Eagles got a point, we got a few goals and points, and then West Coast scored the last 4 points of the match, quite a bit like the Hawks.
Geez I still love that match, it's hard as nails.
The only conclusion that I have is that it was all mental. There weren't any fitness issues for why we played so poorly since we had no injury issues throughout the week, only some concern over Reid. Maybe not so much in recent times, but the Swans do have a track record of shooting themselves in the foot whenever they have the favourite tag applied to them against good opposition, or at least they did back in the Roos days. I think earlier this season we weren't at all good for very obvious reasons, and unfortunately it came back to haunt us at the wrong time. Just smacks of overconfidence and I couldn't really understand why the * they didn't just go back to basics. Maybe they had slow juice or constipation medicine before the game.
Are we that poor? No. Are Hawthorn that good? Maybe, but we certainly made them look like the best ever team to grace a football field, a pretty easy achievement when we could have made Melbourne look like a top 4 side last Saturday.
I likened their performance to Arsenal vs Liverpool earlier this year when Arsenal were already 4-0 down 20 minutes in (http://www.livegoals.com/gamecenter/liverpool-vs-arsenal-08-02-2014), or Arsenal vs Chelsea where they were thumped well and truly, 6-0 (http://www.livegoals.com/gamecenter/chelsea-vs-arsenal-22-03-2014). In those games Arsenal weren't favourites, not even near it, but were expected to draw or go down by 1 goal.
Squiggle got it wrong, but it wasn't because of the algorithm, because no one expected that the Swans wouldn't turn up and try.
- Feb 9, 2011
- 5,996
- 6,485
- AFL Club
- North Melbourne
- Other Teams
- Boston CelticsRedSoxBruins
So North get thumped in a prelim = Norf are s**t
Sydney get pumped in the GF = They didn't turn up?
Note, not a shot at you FS or the Squiggle. Just find it amusing that most deride North as being lucky to be there and Sydney are given a pass for 'not turning up' but being equally blown out of the water.
When Luke Ablett kicked poorly out from the backline only to hit Cousins on the chest, and for that kick to sail back over his head for a goal, that put the Swans about 9-10 points down, and surely all the momentum the Eagles way. (They had outscored the Swans by 6 goals from half time up to that point). It was an astounding way to launch a Premiership-winning burst from for the Swans, many sides would have just about lost their mojo...Luke Ablett was a very relieved Premiership player i'm sure!
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,236
That's fair enough, though, since Sydney had a shocker after an impressive season, while the Roos made it to the prelim by the skin of their teeth. They deserved to be there, but after finishing 6th in the H&A and pulling out narrow escapes in their EF and SF, you'd have to say only just. If North had had Sydney's form going into that prelim, we'd probably say they just underpeformed on the day as well.
Everyone keeps saying The Squiggle got it the GF wrong. I thought The Squiggle predicted a Hawthorn win (albeit with a smaller margin with home ground advantage factored in - something Final Siren tells us shouldn't really apply to GF's though). I thought it was just Final Sirens other models and predictors that suggested a Sydney win. But the main Squiggle predicted a Hawk win. Is that right?
(stands on table..."Oh Squiggle, my Squiggle)
(stands on table..."Oh Squiggle, my Squiggle)
Last edited:
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,238
The algorithm I somewhat accidentally ended up using for Squiggle 2014 did predict a Hawks win, yes, but only because it awarded home state advantage, which (a) certainly did not change the result, and (b) doesn't usually seem to exist in Grand Finals. So it's a correct tip but not one that came from any real insight. And I didn't recommend it in this thread.
The Grand Final is the hardest game for a computer model to tip. In the regular season, a computer can effortlessly keep track of every relevant stat from every game, and compare to hundreds/thousands of cases from the recent past. That's hard for humans to do; we're smarter than computers but we tend to generalize, or miss things, or overrate certain factors, or be affected by our biases. Come Grand Final time, though, we only need to assess two teams and a handful of recent games. There's no huge historical data set to wade through (assuming that Grand Finals are Not Like Other Games, which seems to be true): we only get one a year. You could have the best computer model for tipping GFs in the world and still be unsure if it was any good because the sample size is so small.
crazywildhorse
Brownlow Medallist
2005 was another example of St Kilda doing everything possible to somehow not win the flag.
Sydney moved the ball OK on the few occasions they caught hawthorn on the rebound, but hawthorn suppressed their game plan for most of the game.
In a away it shows the hawks had respect for sydneys potential to come back, they said as much
Bar is rised though. Neither team will repeat with a repeat of last year
Final Siren . Have you applied the squiggle to next years fixtures yet? Sorry if I've missed a new thread.
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,242
No new thread yet, although I'll probably start a Squiggle 2015 one at some point!
Fixture: I'm on holidays in Perth at the moment. But I am curious as to how it will look.
One thing I'd like to do is rate who's got the hardest draw. I've seen two goes at this (The Age and Champion Data) but neither make that much sense to me. The Age in particular is just bizarre, saying that playing the Bulldogs is 500% harder than playing St Kilda, and 6-day breaks are bad regardless of the opposition's break, despite there being no statistical basis for this, among other oddities.
Based on Squiggle logic (TM), I would look at two things only:
1. Strength of opposition in double-up games
+
2. Where you play games against interstate opponents
According to Squiggle, interstate advantage is worth 12 points. The difference between playing an interstate opponent at home compared to away is 24 points. Four goals! That's massive.
It seems to me the ideal fixture for a Victorian club would be playing all non-Vic opponents in Melbourne, and only once, and then having double-ups against all the bottom Melbourne teams. That way you get eight games with home state advantage, and none with a disadvantage.
Of course, this doesn't happen; everyone has to travel. But from my quick glance at the fixture, the Demons may have something pretty close to this.
With a little more sophistication, we should look at which games may be close enough for interstate home advantage to make a difference. For example, Richmond vs Gold Coast is probably close enough for home state advantage to matter, while Melbourne vs Sydney isn't. So if you're Melbourne, you would probably rather take your likely loss against the Swans up in Sydney, since you have to travel at least a few times, and play the Giants in Victoria, where you can beat them.
Thanks for your reply. I look forward to the update. Regarding 6 day breaks, Fremantle football Club have made it clear that they don't like consecutive 6 day breaks, or 6 day breaks before travelling. I would think they have evidence to back that up? I think Sydney benefitted most from opposition team's short breaks, didn't they? Geelong's smashing was at the end of two six day breaks ( plus travel across the country)
At least this year it seems fairer with who gets the games following the bye. Might not be so much of an issue.
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,244
Freo may have evidence that 6-day breaks hurt, and it makes intuitive sense, but last I checked, teams coming off 6-day breaks win 50% of games. There needs to be some good evidence before I start basing models on it. Otherwise it's not stats, it's feels.
Even if shorter breaks do matter, it's clearly silly to consider 6-day breaks without also looking at the opposition's break. The Age model says that if two teams face each other, both coming off a 6-day break, they both have a harder draw. But they can't both be more likely to lose.
There is a good write-up here that suggests 8-day breaks vs 6-day breaks offer a good advantage:
http://www.theroar.com.au/2014/10/30/rts-2015-afl-fixture-breaks-matter/
It also says teams tend to lose after the bye when they play a team that didn't also have a bye.
yellowandblack21
All Australian
But this seems to be the opposite when it comes to finals. Invariably, the team with the semi-final week off wins their prelim.
Not sure how you'd reliably build that into your model though...
treefingers
Cancelled
- Jan 20, 2008
- 6,514
- 8,395
- AFL Club
- Hawthorn
Usually because the team with the week off is higher ranked, and therefore, most likely a higher quality team than the team who played the semi final.
threenewpadlocks
Brownlow Medallist
Would NBA Basketball work?
Similar scorelines to AFL
The main problem that I can see is that as games are generally close, you would have the opposite effect of the NFL. All the teams would be really close together.
- Aug 18, 2009
- 4,229
- 17,495
- AFL Club
- Richmond
- Thread starter
- #2,248
Oh, it's a long and lonely time, the off-season. To make it a little more bearable, here is a 2015 Squiggle preview!
Basically the 2015 Interactive Squiggle is online and has a few new features.
First of all, it's the usual squiggle chart, with 2015 starting positions. You may notice you can hover over a team on the chart to get a handy diagonal line, which is meant to make it easier to compare teams.
The season predictor is up and running, pencilling in a Hawthorn three-peat.
But also! Here is where it gets sexy. You can now drag teams to different positions on the chart, click FORECAST, then GENERATE a season prediction based on their new spots. That's right, you can now correct blatant squiggle errors with your own superior insights! You even get a permalink for sharing with others - e.g. see this Tiger Utopia.
This is still a work in progress, so please let me know of any problems or suggestions for how things could work better!
Basically the 2015 Interactive Squiggle is online and has a few new features.
First of all, it's the usual squiggle chart, with 2015 starting positions. You may notice you can hover over a team on the chart to get a handy diagonal line, which is meant to make it easier to compare teams.
The season predictor is up and running, pencilling in a Hawthorn three-peat.
But also! Here is where it gets sexy. You can now drag teams to different positions on the chart, click FORECAST, then GENERATE a season prediction based on their new spots. That's right, you can now correct blatant squiggle errors with your own superior insights! You even get a permalink for sharing with others - e.g. see this Tiger Utopia.
This is still a work in progress, so please let me know of any problems or suggestions for how things could work better!
Pretty sure a few of us asked for Squiggles to be integrated into our daily lives. I think you've succeeded. I tip my hat to you sir.
If only...
Last edited:
Similar threads
