Certified Legendary Thread Race for the flag, in squiggly lines

[Final Siren]

There have been a lot of upsets in the past decade that have been hard to forecast even with the benefit of hindsight.

Yes, which raises juicy questions of what the heck happened! 2014 is an interesting year because the Hawks and Swans seemed very evenly matched going in, with hardly anyone predicting a blowout. And when a team wins by 10 goals, you have to think that's no accident. So it should have been predictable and the question is what factors should you have considered to predict it.

 

[SJ]

Yes, which raises juicy questions of what the heck happened! 2014 is an interesting year because the Hawks and Swans seemed very evenly matched going in, with hardly anyone predicting a blowout. And when a team wins by 10 goals, you have to think that's no accident. So it should have been predictable and the question is what factors should you have considered to predict it.

Not everything is predictable though. We don't have access to all contributing factors, nor can we ever predict randomness.

And also I think in grand finals, margins don't mean as much. If the same game was played 100 times, it's entirely possible that Sydney would have won more than half.

 

[Hawkk]

Hawthorn had won 4 of the previous 5 games against the Swans leading up to that GF (their only lose was a game that they were decimated by injury)

The Hawks won the previous clash between the clubs (round 18) at their home ground, a ground that happened to be the venue for the Grand Final

This is of course retrospective though...

 

[Final Siren]

Not everything is predictable though. We don't have access to all contributing factors, nor can we ever predict randomness.

And also I think in grand finals, margins don't mean as much. If the same game was played 100 times, it's entirely possible that Sydney would have won more than half.

That would mean that what we saw was the 1% outlier, right? And on the other 99 alternate-reality Earths, the game was closer, or the Swans won. Yeah, it's possible, but of course first you go looking for explanations that rely on less randomness.

Clearly not everything is predictable, especially with limited inputs, but a 10+ goal deviation from prediction in a match where both teams are 100% focused on the result is pretty alarming to me. For that, I'd want to hear that Longmire & Co. were threatened by match-fixing bandits before I'd feel comfortable with my existing model.

Fremantle have been very interesting to me this year, because the squiggle says they're a pretty average team plus a small dollop of luck/randomness. That is, even though they were the minor premier, their results in the second half of the season were well within a reasonable margin of error for an 8th-or-thereabouts team. I really wonder how different seasons would be if we got to watch them play them out in a few alternate-reality Earths. Do North Melbourne more commonly convert close losses to close victories in 2013 and make the prelims instead of missing finals? Do Port do the same thing this year, while North miss?

It would be a bit dispiriting if this were true, because we all want to believe that there's no such thing as luck in football, the best team prevails on the day, etc etc. But yeah. I wonder.

 

[Final Siren]

That would mean that what we saw was the 1% outlier, right? And on the other 99 alternate-reality Earths, the game was closer, or the Swans won. Yeah, it's possible, but of course first you go looking for explanations that rely on less randomness.

Clearly not everything is predictable, especially with limited inputs, but a 10+ goal deviation from prediction in a match where both teams are 100% focused on the result is pretty alarming to me. For that, I'd want to hear that Longmire & Co. were threatened by match-fixing bandits before I'd feel comfortable with my existing model.

Fremantle have been very interesting to me this year, because the squiggle says they're a pretty average team plus a small dollop of luck/randomness. That is, even though they were the minor premier, their results in the second half of the season were well within a reasonable margin of error for an 8th-or-thereabouts team. I really wonder how different seasons would be if we got to watch them play them out in a few alternate-reality Earths. Do North Melbourne more commonly convert close losses to close victories in 2013 and make the prelims instead of missing finals? Do Port do the same thing this year, while North miss?

It would be a bit dispiriting if this were true, because we all want to believe that there's no such thing as luck in football, the best team prevails on the day, etc etc. But yeah. I wonder.

Heh, I just noticed that if North had squeaked into 8th in 2013, they would have played Richmond in the first Elimination Final (instead of Carlton), and probably won that, just like this year, then played Sydney in the semi-, just like this year. This is the season North should have had two years ago.

 

[The Passenger]

Yes, which raises juicy questions of what the heck happened! 2014 is an interesting year because the Hawks and Swans seemed very evenly matched going in, with hardly anyone predicting a blowout. And when a team wins by 10 goals, you have to think that's no accident. So it should have been predictable and the question is what factors should you have considered to predict it.

IMO last year was a classic case of one team getting on top, momentum just snow balling and the other team just not being able to stop it. I said to my Swans supporting best mate that I wouldn't read too much into that match and the margin. If the GF was a best of 3 it was quite conceivable the Swans come back and win the next two. The competition at the top is so even that you only need a 2 or 3 percent swing in on-ground performance from a team and the margin can easily blow out like last years GF.

Since I've been knowledgeably following football (20 years) I can't remember a better team at cutting up an opposition when they are on top as Hawthorn are. They are the kings of scoreboard pressure IMO and it's so important, particularly in big games, to put points on the scoreboard when you have momentum. Hawks do it better than anyone I can remember. That's what has hurt Freo over the last 3-4 years, and it hurt North in the first quarter against us on Saturday night. North should have been up 6+ goals at quarter time. Over 2015, we've been just as good as Hawks as matching them in that area. We may not have been doing consistently year in year out, but over 2015 we have certainly been brilliant at punishing teams and it's crucial. It really kills a team to have almost all the play in a quarter and only win it by a couple of goals.

I said in the game day thread if either team doesn't turn up, or is slightly off their game, the result could really blow out this weekend. This is much more of a concern for us then the Hawks. Rising teams like the Eagles tend to have off days and big finals bring out the nerves. I'm really hoping the PF scare was just what we needed to snap out of any potential malaise heading into the GF. The Hawks haven't too many off days over the last few years, but they have been more susceptible than normal to them over the last two months, and we saw in the QF that if we get on top we can put them to the sword with the margin in that match rising to 50 points at one stage.

But if both teams turn up and give it their best then it should be a pretty close contest with plenty of points scored which will be great for neutral fans, but perhaps not great for the tickers of us and Hawks supporters.

 

[SJ]

That would mean that what we saw was the 1% outlier, right? And on the other 99 alternate-reality Earths, the game was closer, or the Swans won. Yeah, it's possible, but of course first you go looking for explanations that rely on less randomness.

What was the squiggle's mean error in 2014?

 

[Final Siren]

So just to clarify the only Grand Final predictor that uses OFFDEF4G-95:79 is the 2015 tip?

Right, I've never gotten around to regenerating all those past years, so they're still ISTATE-91:12 from start to finish.

I will probably regenerate them sometime over the off-season.

 

[Final Siren]

What was the squiggle's mean error in 2014?

By margin? 22.8 points.

Edit: You're going to ask for a standard deviation next, aren't you. I would have to calculate that.

 

[Hawkk]

Right, I've never gotten around to regenerating all those past years, so they're still ISTATE-91:12 from start to finish.

I will probably regenerate them sometime over the off-season.

I notice that the full season tips / actual results for season 2015 are no longer accessible...

I wanted to do a 5 to 10 round breakdown identifying games where the Hawks and Eagles have broken / failed to meet the predicted line...

All I could find are the finals tips in this thread...

QF
Eagles 96 def Hawks 64 (squiggle tipped Eagles 93-84) (+23)

SF
Hawks 134 def Crows 61 (squiggle tipped Hawks 112-81) (+42)

PF
Hawks 94 def Dockers 67 (squiggle tipped Hawks 87-74) (+12)
Eagles 80 def North 55 (squiggle tipped Eagles 101-74) (-1)

Interestingly aside from the Eagles v North (which was subject to deplorable kicking at goal) and the Hawks score against the Eagles, scoring has been pretty much in line with prediction

 

[Richo83]

I think the grand final is the big outlier of the year though. Both teams are given a week's break during the finals, it's not played on the best team's ground like the rest of the finals are, and favours Victorian teams, meaning upsets are less common. Teams plan differently for the final compared to other matches given it's the last match. And the playing field is reduced by the fact that it's a one-off match, not a best of seven series like baseball, and not over 38 matches like European leagues do. Just look at Port Adelaide, in 2007 they had the biggest loss they had ever suffered in AFL history, and it came in a final where they were the second best in the league. That makes no sense, but in a sport which is high scoring and in such an unusual format, it makes some form of sense.

 

[The Punter]

it's not played on the best team's ground like the rest of the finals are, and favours Victorian teams, meaning upsets are less common.

1991 Hawthorn d West Coast
1992 West Coast d Geelong
1994 West Coast d Geelong
1996 North Melbourne d Sydney
1997 Adelaide d St Kilda
1998 Adelaide d North Melbourne
2001 Brisbane d Essendon
2002 Brisbane d Collingwood
2003 Brisbane d Collingwood
2007 Geelong d Port Adelaide
2012 Sydney d Hawthorn
2013 Hawthorn d Fremantle
2014 Hawthorn d Sydney

The data doesn't bear that out. Also, out of the eight times (out of 13) the non-Victorian team has beaten the Victorian team in the Grand Final, only twice have they finished higher on the ladder than their opponent (1994 & 2002).

 

[Final Siren]

I notice that the full season tips / actual results for season 2015 are no longer accessible...

I wanted to do a 5 to 10 round breakdown identifying games where the Hawks and Eagles have broken / failed to meet the predicted line...

All I could find are the finals tips in this thread...

QF
Eagles 96 def Hawks 64 (squiggle tipped Eagles 93-84) (+23)

SF
Hawks 134 def Crows 61 (squiggle tipped Hawks 112-81) (+42)

PF
Hawks 94 def Dockers 67 (squiggle tipped Hawks 87-74) (+12)
Eagles 80 def North 55 (squiggle tipped Eagles 101-74) (-1)

Interestingly aside from the Eagles v North (which was subject to deplorable kicking at goal) and the Hawks score against the Eagles, scoring has been pretty much in line with prediction

I just added a "[+] History..." button to the TIPS tab so you can see this kind of thing.

 

[Hawkk]

Just reading the Grand Final myths thread on the main board I thought I would have a look at the impact of the 4 game finals series using the squiggle performance in forecasting the Grand Final result (where the two competing clubs have played an uneven number of finals games)

For ease of reference I extended the analysis pre final 8 (1994 onwards) to the creation of the final 6 (1991 - 1993) and final 5 (1972 - 1990)

Top 8 (QF, SF, PF, GF (4 games) vs. QF, PF, GF (3 games))
1994 ISTATE:91-12 tipped Eagles by 22 (Eagles (rested) won by 80pts in the GF)
1997 ISTATE:91-12 tipped St Kilda by 22 (Crows won by 31 pts in the GF)
1998 ISTATE:91-12 tipped North by 4 (Crows won by 35 pts in the GF)
1999 ISTATE:91-12 tipped North by 9 (North (rested) won by 35 pts in the GF)
2003 ISTATE:91-12 tipped Lions by 17 (Lions won by 50 pts in the GF)
2005 ISTATE:91-12 tipped Swans by 25 (Swans won by 4 pts in the GF)
2006 ISTATE:91-12 tipped Swans by 8 (Eagles won by 1 pt in the GF)

Top 6 (QF, SF, PF, GF (4 games) vs. QF, SF, GF (3 games))
1991 ISTATE:91-12 tipped Hawks by 9 (Hawks (rested) won by 53 pts in the GF)
1992 ISTATE:91-12 tipped Cats by 15 (Eagles (rested) won by 28 pts in the GF)
1993 ISTATE:91-12 tipped Dons by 4 (Dons won by 44 pts in the GF)

Top 5 (QF, QF, SF, GF (4 games) vs. SF, PF, GF (3 games))
1990 ISTATE:91-12 tipped Pies by 16 (Pies won by 48 pts in the GF)

Top 5 (QF/EF, SF, PF, GF (4 games) vs. SF, GF (2 games)
1989 ISTATE:91-12 tipped Hawks by 12 (Hawks (rested) won by 12 pts in the GF)
1988 ISTATE:91-12 tipped Hawks by 32 (Hawks (rested) won by 96 pts in the GF)
1987 ISTATE:91-12 tipped Hawks by 4 (Blues (rested) won by 31 pts in the GF)
1985 ISTATE:91-12 tipped Dons by 1 (Dons (rested) won by 78 pts in the GF)
1982 ISTATE:91-12 tipped Blues by 9 (Blues won by 18 pts in the GF)
1981 ISTATE:91-12 tipped Blues by 12 (Blues (rested) won by 20 pts in the GF)
1979 ISTATE:91-12 tipped Blues by 5 (Blues (rested) won by 7 pts in the GF)
1975 ISTATE:91-12 tipped North by 0 (North won by 55 pts in the GF)
1974 ISTATE:91-12 tipped Tigers by 1 (Tigers (rested) won by 41 pts in the GF)

Top 5 (QF, SF, GF (3 games) vs. SF, PF GF (3 games)
1986 ISTATE:91-12 tipped Hawks by 8 (Hawks won by 42 pts in the GF)
1984 ISTATE:91-12 tipped Dons by 11 (Dons won by 24 pts in the GF)
1978 ISTATE:91-12 tipped Hawks by 1 (Hawks won by 18 pts in the GF)
1972 ISTATE:91-12 tipped Tigers by 5 (Blues won by 27 pts in the GF)

Top 5 (QF/EF, SF, PF, GF (4 games) vs. QF, SF, GF (3 games)
1983 ISTATE:91-12 tipped Hawks by 6 (Hawks (rested) won by 83 pts in the GF)
1980 ISTATE:91-12 tipped Tigers by 8 (Tigers (rested) won by 81 pts in the GF)
1976 ISTATE:91-12 tipped North by 4 (Hawks (rested) won by 30 pts in the GF)
1973 ISTATE:91-12 tipped Blues by 4 (Tigers won by 30 pts in the GF)

Top 5 (QF, SF, PF, GF, GF (5 games) vs. SF, GF, GF (3 games)
1977 ISTATE:91-12 tipped North by 6 + 5 (North (rested) won replay by 27 pts in the GF)

Ignore the Top 5 (5 games v 3 games) and (4 games v 2 games) and the rested team beats the squiggle performance line 9 times out of 17 games (53% of time)

That probably goes against conventional understanding...

 

[Hawkk]

The 1977 North Melbourne side is probably the best club win of that lot. They won a 5 game final series and actually looked fitter in the fifth game...

 

[The Punter]

Of course a purely statistical analysis removes the human element.

After the 1977 draw, Collingwood trained as usual, which was the usual Tommy Hafey treatment - bloody hard. Barassi decided North should have a very light week on the track.

In 1993 Essendon had figured out the Carlton centre bounce set up, but Sheedy didn't want to let Carlton know that until it was completely necessary. Essendon also played a bunch of blokes in that first final against Carlton who wouldn't play in the GF.

 

[Brishawk]

IMO last year was a classic case of one team getting on top, momentum just snow balling and the other team just not being able to stop it. I said to my Swans supporting best mate that I wouldn't read too much into that match and the margin. If the GF was a best of 3 it was quite conceivable the Swans come back and win the next two. The competition at the top is so even that you only need a 2 or 3 percent swing in on-ground performance from a team and the margin can easily blow out like last years GF.

Since I've been knowledgeably following football (20 years) I can't remember a better team at cutting up an opposition when they are on top as Hawthorn are. They are the kings of scoreboard pressure IMO and it's so important, particularly in big games, to put points on the scoreboard when you have momentum. Hawks do it better than anyone I can remember. That's what has hurt Freo over the last 3-4 years, and it hurt North in the first quarter against us on Saturday night. North should have been up 6+ goals at quarter time. Over 2015, we've been just as good as Hawks as matching them in that area. We may not have been doing consistently year in year out, but over 2015 we have certainly been brilliant at punishing teams and it's crucial. It really kills a team to have almost all the play in a quarter and only win it by a couple of goals.

I said in the game day thread if either team doesn't turn up, or is slightly off their game, the result could really blow out this weekend. This is much more of a concern for us then the Hawks. Rising teams like the Eagles tend to have off days and big finals bring out the nerves. I'm really hoping the PF scare was just what we needed to snap out of any potential malaise heading into the GF. The Hawks haven't too many off days over the last few years, but they have been more susceptible than normal to them over the last two months, and we saw in the QF that if we get on top we can put them to the sword with the margin in that match rising to 50 points at one stage.

But if both teams turn up and give it their best then it should be a pretty close contest with plenty of points scored which will be great for neutral fans, but perhaps not great for the tickers of us and Hawks supporters.

I think everyone was dismissive of Hawthorn's record against the swans since 2012. Hawks had an average winning margin against the swans of 28 points leading into the grand final (excluding the loss by 19 points earlier in the season). The only final played between the two teams in 2013 was a 54 point margin to the hawks. The average winning margin by Hawthorn has climbed to 44 since 2012 (excluding both losses). That is a not an anomaly.

 

[Roby]

I think everyone was dismissive of Hawthorn's record against the swans since 2012. Hawks had an average winning margin against the swans of 28 points leading into the grand final (excluding the loss by 19 points earlier in the season). The only final played between the two teams in 2013 was a 54 point margin to the hawks. The average winning margin by Hawthorn has climbed to 44 since 2012 (excluding both losses). That is a not an anomaly.

That's one thing that usually is the best sign to premiership success. Premiership teams tend not to get smashed, not even one game.

Look at the Hawks in recent years, they may have lost games but they have never played a poor game, a poor game where they are comprehensively beaten.

In 2014, Sydney played a game against North at the SCG in the wet, where they lost by 43 points. The Hawks worst loss in that same year was 20 points.

In 2013 Fremantle had 3 losses above 40+ points, two against finals challengers Hawthorn and Geelong, and the other in mitigating circumstances, but none the less it was 71 points. The Hawks did have one poor loss in rnd 19 against the Tigers where they were still in the game till 3/4 time and then they blew them of the park in the last quarter, but there worst next loss was 10 points.

In 2012, Hawks had their third worst lost since in the last 8 year year (and their worst loss since then), against the Tigers again in round 9 in a very wet game, they also lost by 37 points to Sydney at home. Sydney in contrast that year, their worst loss was 36 point defeat to Geelong at Kardinia, in a game they kicked very poorly in front of goal. Their other losses were all under 30 points.

In 2011, same thing. Collingwood only lost three games that year but it was against their grand final opponent in Geelong every time, and while it was a dead rubber in the last round they still were beaten by nearly 100 points. Geelong in contrast that year their worst loss was by 13 points.

2010, same thing. Pies worst loss 36 points, St Kilda 61 points. They even lost to the Pies by 48 points half way through the season.

2009 is the exception, but there is bradbury there if you have ever seen. Everything went right for Geelong in 09 Grand Final, St Kilda won the first three quarters and were the first only team ever to do so and not win the Grand Final.

Even 2008 the Hawks worst loss was 32 point loss to the Bulldogs, but the Cats in round 9 suffered a 86 point loss to Collingwood.

 

[Roby]

This year West Coast have has had 57 point loss (do we count that game?) none the less they had it, Hawthorn's worst loss was in the qualifying final by 32 points.

Fremantle are very defensive but they still had 72 and 69 point losses this year. North, 55, 60, 73 & 77. Sydney 52 & 82. Adelaide 52, 56 & 74 (Semi-Final). Dogs two losses 70 and above.

Richmond worst loss was 36, so I don't know what is happening with them. I'm guessing if Richmond had beaten North as they should have, then they would have beaten Sydney as well. And then who knows after that, given the umpiring in that game, North played better, maybe they would have beaten West Coast as well. Richmond are just odd.

 

[Final Siren]

Finals Week 3, 2015

Animated!

Not much movement for Fremantle and Hawthorn, while the low-scoring Eagles/Roos game drove both teams down and right.

That means it's a Hawk tip for the Grand Final!

SQUIGGLE:

Hawthorn 94 - 82 West Coast

This tip comes courtesy of Flagpole. (ISTATE-91:12, the algorithm used for the rest of the season, would tip similarly, 92-77, as it has the teams rated closer together but gives Hawthorn home game advantage.)

Speaking of which, Flagpole now looks like this:

Flagpole actually did really well this year. Here's the chart of flagpole heights throughout the season:

So that's a nice early identification of our eventual Grand Finalists. If the Hawks get up, it's basically a faultless season. I also noticed it beat the squiggle in regular tipping (it has 151 tips to this point vs squiggle on 145). Its error margin isn't nearly as good, but still, that's a top effort.

More details about Flagpole workings below, just in case you care. But the thing of note is that Flagpole is biased towards attacking teams, and 2015 is another year of evidence against the idea that "defence wins flags."

Not only did the two most attacking teams wind up in the GF, but we had three defensive specialists in the top five, who exited the finals as quickly as humanly possible.

Why Flagpole

Flagpole came about because this year really exposed the fragility of predicting premiers by tipping finals game-by-game. As squiggle regulars know, the Home & Away Ladder Predictor doesn't simply tally up tips, but rather adds together the probabilities of wins, which allows it to account for the likelihood that some tips will be wrong. Without this, favourites would be predicted to win every game, and wooden spooners to lose the lot, which doesn't happen in real life.

But the finals predictor, using straight tips, was producing very different outcomes week-to-week based on knife-edge results, like whether a team finished 2nd or 3rd and got a home final. And while it's interesting to consider butterfly effects like that, it wasn't accurately expressing teams' flag chances as the squiggle saw them.

In particular, from eye-balling the chart, you could clearly see Hawthorn as the best-placed team, since they spent all year surrounded by premiership cups. Only the Eagles even got close. But the finals predictor often had the Hawks dropping an away final, since they were likely to miss top 2, and therefore bowing out early.

So Flagpole was designed to remove the fragility of single-game tipping and provide a more robust rating of where each team stands with regard to premiership-winning form.

How To Make Your Own Flagpole (a.k.a. OFFDEF4G-95:79)

1. Assign every team an ATTACK score of 50 and a DEFENCE score of 50
2. Look up the results of the first game of last year
3. Predict scores using the formula: TEAM A PREDICTED SCORE = 79 * TEAM A ATTACK / TEAM B DEFENCE
4. Compare predicted scores to actual scores by using the same relationship described above, so that TEAM A HYPOTHETICAL ATTACK = TEAM A ACTUAL SCORE * TEAM B DEFENCE / 79 and TEAM A HYPOTHETICAL DEFENCE = 79 * TEAM A ATTACK / TEAM B ACTUAL SCORE.
5. Here is the unique thing about OFFDEF4: square and double the HYPOTHETICAL ATTACK score. That is: NEW HYPOTHETICAL ATTACK = HYPOTHETICAL ATTACK * HYPOTHETICAL ATTACK * 2.
6. Assign new ATTACK and DEFENCE scores to each team, where 95% of the new score is the old score and 5% is the new score. That is: TEAM A NEW ATTACK = (0.95 * TEAM A OLD ATTACK) + (0.05 * TEAM A NEW HYPOTHETICAL ATTACK). Ditto for DEFENCE.
7. Repeat steps #2-#6 for every game.

That's it! Just one other thing: because this algorithm generates a lot of inflation over a season (which is good, I believe, because it means more more weighting for finals performances), its tips are only good as margins. If you want it to predict absolute scores, you need to increasingly scale back its predicted numbers, or else you get Grand Final tips where both teams are tipped to score well over 100.

Note: I just realized I completely buggered up earlier charts in this thread for flagpole heights across a season, because I was only charting ATTACK scores, not combined scores. Sorry about that. Those charts are useless.

 

[Hawkk]

Do you have predictors that are forecasting an Eagles win?

I cant believe how little love they are getting...

 

[floki]

Have I ever told you squiggle, I love you, I love you with all my heart.

 

[Final Siren]

Do you have predictors that are forecasting an Eagles win?

I cant believe how little love they are getting...

Well, it's that thing where 100% of rational experts will tip a 60/40 favourite. But that doesn't mean it's 100% likely to win.

About 90% of my algorithms are tipping the Hawks, 10% Eagles.

Curiously, one of the Eagles boosters is OFFDEF-75, which was my GF tipper last year, until the Hawks disgraced it. It isn't impressed with the Hawks' late-season form, while it loves how the Eagles' last three games involved keeping St Kilda to 30pts, Hawthorn to 64pts, and North to 55pts.

 

[The Lemon]

Do you have predictors that are forecasting an Eagles win?

I cant believe how little love they are getting...

Yep i want to know how many predictors have WCE...because they have this win

 

[The Lemon]

Curiously, one of the Eagles boosters is OFFDEF-75, which was my GF tipper last year, until the Hawks disgraced it. It isn't impressed with the Hawks' late-season form, while it loves how the Eagles kept St Kilda to 30pts in Round 23 and North to 55pts last weekend.

So one result is enough to stop it?

I think the reasons it gives are scary