Certified Legendary Thread Race for the flag, in squiggly lines

[Final Siren]

The hawks sure do know how to crush teams don't they!

Final Siren with the Syd - Hawks game coming. What results would most impact the squiggle.

Does the squiggle react to what we perceive as a key fixture as much as the football public might? e.g.. close loss either way no big deal. Sydney win by 4 + Goals people will be talking about hawthorns long term hopes given they go back 3 games etc. Sydney lose and they are neck and neck with the hawks on the ladder..

Good question. I think compared to the football public, the squiggle cares less about "key games" and more about results vs bad teams. It expects different results depending on the strength of the opposition, but it will learn just as much about a team from how it performs against the wooden spooner as it does against the defending premier.

My opinion is that we humans tend to look for reasons to dismiss games too easily, because we can't keep every match in our head at once. We need to reduce the data down to a manageable set. So we ignore some results that we're not sure how to interpret, or generalize a set of games into something like, "West Coast always smash bad teams." Humans are pretty good at generalizing like this, and at finding patterns, but not perfect.

There would have been occasions in the past when the squiggle has gotten over-excited over a team's performance against bad opposition... I think that does happen. Probably just not as much as people might think.

As for margins, for the purposes of rating form, the squiggle cares about the margin, but not especially about who wins. That is, there's no real difference between a 1-point win and a 1-point loss, as far as the squiggle is concerned; either way, it's just a close game. Whereas the football public would say there is a big difference, and the winner proved they can close out tight contests, etc etc.

But who wins matters a lot in the ladder predictor! As an example, let's look at the GWS vs Adelaide game coming up this weekend. The squiggle is tipping GWS by 3 points. Because it's a probabilistic ladder, as discussed earlier, it works by awarding GWS 0.55 wins for this match and Adelaide 0.45 wins. By the end of the year, as it happens, both teams have almost exactly the same number of predicted wins: 13.14 for GWS and 13.13 for Adelaide. These are both rounded off to 13 wins.

Now let's say that this weekend plays out exactly according to the squiggle tip: GWS win 93 to 90. Neither team will move at all on the squiggle chart. But the ladder predictor will change, because it's now able to eliminate the chance that its tip was wrong. GWS will change from 0.55 wins to 1 win for Round 8, and Adelaide will change from 0.45 wins to 0 wins. Their final tally of predicted wins will likewise change: GWS will gain 0.45 wins to go from 13.14 to 13.69 (rounded off to 14) and Adelaide will drop 0.45 wins to go from 13.13 to 12.68 (rounded off to 13).

There's often a significant difference between the squiggle's ranking of teams on form and where it predicts them to finish at the end of the year. For example, West Coast are currently ranked 4th on form, but from 5-2 with a good fixture, they're predicted to finish 3rd, with 16 wins. Port are ranked 5th on form, only one below West Coast, but after a 3-4 start and with a tough fixture still to come, are predicted to finish 8th with 13 wins, five ladder rungs lower. (And then beat the Crows, 5th, in an Elimination Final, because they're still ahead on form.)

 

[The Swans Blog]

INCREDIBLY ACCURATE

No, what you really should be asking is why it's assessing them this way. The answer is:

• The Pies started 2015 ranked quite low on the back of a series of abysmal late-2014 performances. (You can say this was due to injury, though.)
  
• They beat Brisbane (just), St Kilda (comprehensively), Essendon (20 pts), and Carlton (comprehensively). The squiggle rates three of those as terrible teams and the Bombers as mediocre.
• They lost to Adelaide (by 27 in Melbourne), Geelong (by 41), and Richmond (just). Adelaide is a mid-to-high table team, Geelong is mid-table, and Richmond is low-to-mid.

So essentially the Pies haven't played anyone really good, and haven't been great against most of the teams they have played.

Squiggles have been the most accurate model so far this season when it comes to predicting margins. Season average margin has been pretty good and last weekend was really good, tipping 5 games under 12 points.

 

[woosha24]

It is quite possible we are just lost...

Quite possibly.

WCE symbol: "What the f*ck? Fremantle? Sydney? This isnt the Bottom 6. The f*cken hell are we!?"

 

[The Falcon Strike]

If Sydney and West Coast switched places that would be one hell of a way to kick off the finals series.

The way it is now though lends itself to cracking prelims...

 

[Illinois Nazi]

May not be a popular question but;

Does the squiggle overrate thrashings of poor teams?

The only teams that West Coast have beaten that are worth anything are Port and GWS - GWS who are hard to read (young team travelling across the country) and Port (having a miserable season).

This week West Coast have the Saints - could get nasty again.

It's not so much thrashings as keeping teams to a very low score that the squiggle possibly over-reacts to. (Though obviously those things often go together)

If you win by 60 points, 80-20, you've scored four times your opponent, which is seen as a bigger win than 160-100, which is only 1.6 times your opponent. The 4x win has a more dramatic effect on the squiggle. From the weekend for West Coast it's more the keeping them under 50 points rather than scoring 130 that has driven their movement - hence they moved a long way right but only a little up.

I used to do a mathematical ratings system myself which had a similar issue. I used to find it fun to watch a team get a massive boost from a skewed score line like that, then slowly drift back in subsequent weeks to their true position. Although the Eagles do seem to be on a sustained mission to the right hand side...

 

[Final Siren]

Squiggles have been the most accurate model so far this season when it comes to predicting margins. Season average margin has been pretty good and last weekend was really good, tipping 5 games under 12 points.

I have 157 algorithms in my main stable now (most are variations of each other) and it's always interesting to see which do well or poorly in a particular year. ISTATE-91:12 is ol' reliable, but with 41 correct tips it's only 86th, even after a great Round 7.

The very worst algorithms are HOMER variants, which is really unusual! Tipping the home team each game is worse than flipping a coin this year: 30 right, 33 wrong.

MOMENTUM, which tips whoever won last week, tie-breaking on margin, is also a terrible strategy this year: 33 right, 30 wrong.

The current star is PILGRIM-96:6, which is less sensitive to recent form, and awards a small home game advantage based on how far teams have to travel. That has 46 tips right (5 more than ISTATE) at 73% accuracy, and is also quite a lot more accurate on margins.

With this many algorithms, one will inevitably leap out in front from pure chance, but I keep an eye on it. PILGRIM is tipping the Crows this weekend over GWS but is otherwise the same as ISTATE.

 

[The Swans Blog]

It's not so much thrashings as keeping teams to a very low score that the squiggle possibly over-reacts to. (Though obviously those things often go together)

If you win by 60 points, 80-20, you've scored four times your opponent, which is seen as a bigger win than 160-100, which is only 1.6 times your opponent. The 4x win has a more dramatic effect on the squiggle. From the weekend for West Coast it's more the keeping them under 50 points rather than scoring 130 that has driven their movement - hence they moved a long way right but only a little up.

I used to do a mathematical ratings system myself which had a similar issue. I used to find it fun to watch a team get a massive boost from a skewed score line like that, then slowly drift back in subsequent weeks to their true position. Although the Eagles do seem to be on a sustained mission to the right hand side...

Collingwood def. Port Adelaide by about 100 points in 2011 right before the finals and they went way off to the right to almost 100 defence. Then of course there's the 1989 game between Geelong & Melbourne, where Geelong won by 69 points, 89 v 20, and moved 25 points to the right.

 

[The Swans Blog]

I have 157 algorithms in my main stable now (most are variations of each other) and it's always interesting to see which do well or poorly in a particular year. ISTATE-91:12 is ol' reliable, but with 41 correct tips it's only 86th, even after a great Round 7.

The very worst algorithms are HOMER variants, which is really unusual! Tipping the home team each game is worse than flipping a coin this year: 30 right, 33 wrong.

MOMENTUM, which tips whoever won last week, tie-breaking on margin, is also a terrible strategy this year: 33 right, 30 wrong.

The current star is PILGRIM-96:6, which is very sensitive to recent form, and awards a small home game advantage based on how far teams have to travel. That has 46 tips right (5 more than ISTATE) at 73% accuracy, and is also quite a lot more accurate on margins.

With this many algorithms, one will inevitably leap out in front from pure chance, but I keep an eye on it. PILGRIM is tipping the Crows weekend over GWS but is otherwise the same as ISTATE.

Any chance you could post the PILGRIM-96:6 alongside ISTATE-91:12? ISTATE is performing quite well and I would like to add it to the model list. I do find it quite interesting that against PR/MR/FMI/FF, Squiggles out-performs when margin is concerned and is close to a full point less average margin. Round 7 was the best round for Squiggles so far this season.

 

[nameless]

Nice to see our team rated, and I'm very optimistic of our team on a whole.
But with injuries, a lot of younger players and a few players in career best form at the moment, I'm not sure we can keep it all together for a full season.

We have a tough patch of about 5-6 games after the bye and I reckon that could knock us around and see us finish around 10th.
Which is where, after losing so many players, I expected us to be abouts.

Be great for us to just make the finals though.

 

[RookiePick]

It's not so much thrashings as keeping teams to a very low score that the squiggle possibly over-reacts to. (Though obviously those things often go together)

If you win by 60 points, 80-20, you've scored four times your opponent, which is seen as a bigger win than 160-100, which is only 1.6 times your opponent. The 4x win has a more dramatic effect on the squiggle. From the weekend for West Coast it's more the keeping them under 50 points rather than scoring 130 that has driven their movement - hence they moved a long way right but only a little up.

I used to do a mathematical ratings system myself which had a similar issue. I used to find it fun to watch a team get a massive boost from a skewed score line like that, then slowly drift back in subsequent weeks to their true position. Although the Eagles do seem to be on a sustained mission to the right hand side...

I was filthy when GC got 4 in the last. I was actually getting a bit excited about how much the eagle was going to zoom if we kept them under 30pts #squigglehasruinedme

 

[Final Siren]

Any chance you could post the PILGRIM-96:6 alongside ISTATE-91:12? ISTATE is performing quite well and I would like to add it to the model list. I do find it quite interesting that against PR/MR/FMI/FF, Squiggles out-performs when margin is concerned and is close to a full point less average margin. Round 7 was the best round for Squiggles so far this season.

PILGRIM hasn't been good enough for long enough. Last year it did 72.0%, the same as ISTATE, with slightly better margin accuracy, but it's done a little worse the three years before that.

ISTATE-91:12 is the proven performer. Here it is vs my current stable over the last 5 years:

Year | Rank | Accuracy \ 2014 | 22nd | 72.0% \ 2013 | 39th | 72.5% \ 2012 | 9th | 77.8% \ 2011 | 32nd | 77.6% \ 2010 | 18th | 68.3%

And over a longer time frame, like more than five years, it's 1st or very close.

 

[lmach]

I was filthy when GC got 4 in the last. I was actually getting a bit excited about how much the eagle was going to zoom if we kept them under 30pts #squigglehasruinedme

Simmo said after the game that he was getting a bit fired up in the box in the second half and apologised to the other coaches. Now we know why.

 

[Macca81]

When will the squiggle project the Eagles up instead of just to the right? 3 out of our last 4 games weve had 115pts or more.

 

[The Falcon Strike]

It's not so much thrashings as keeping teams to a very low score that the squiggle possibly over-reacts to. (Though obviously those things often go together)

Not for us though

When we have towelled teams in the past - we have often kicked 26-30 goals and still conceded 10+.

This year we have kept sides to 6 goals or twice. In fact the GWS game was the second lowest score we have kept a team to in over 20 years.

 

[clogged]

When will the squiggle project the Eagles up instead of just to the right? 3 out of our last 4 games weve had 115pts or more.

When you kick a high score against a team with a good defence.

 

[B4Bear]

I keep thinking that North are top 4 material.
The Squiggle just has us tooling around in the zone of mediocrity.

 

[Strapping Young Lad]

Final Siren I was just looking through the predictions for the entire Home and Away season, and when Hawthorn meet the Saints the predictor says Hawthorn will win yet the ladder for that round records a loss against Hawthorn. I don't know if this anomaly happens for other teams at all - I was only interested in following the predictions through the year for my side.

 

[SJ]

Final Siren I was just looking through the predictions for the entire Home and Away season, and when Hawthorn meet the Saints the predictor says Hawthorn will win yet the ladder for that round records a loss against Hawthorn. I don't know if this anomaly happens for other teams at all - I was only interested in following the predictions through the year for my side.

This gets asked every week:

Season Predictor
This is an estimate of how the ladder is most likely to turn out, on the assumption that the squiggle has correctly assessed the strength of every team and nobody gets much better or worse.

For the home & away season, it uses a probabilistic ladder. That means that instead of tipping individual games and tallying up the tips, it awards a win probability for each game. For example, if the squiggle thinks Hawthorn is 68% likely to beat Collingwood, it will award 0.68 wins to the Hawks and 0.32 to the Pies, increasing both team's tally of "probable wins" by less than 1.

This is because if a team plays 10 games with 60% likelihood of winning each game, we should expect them to win about 6/10—not, as we would get if we tipped each game and tallied up the tips, 10/10. We know that upsets will happen; we just don't know when. A probabilitistic ladder accounts for the likelihood that teams will sometimes unexpectedly win or lose, even though we doesn't know when.

This can look like a bug in the predictor, if you see that a team is predicted to win a match but it doesn't seem to be credited for it. For example, a team might be on "15 (14.7)" wins: that is, 14.7 "probable wins" rounded off to 15. (Rounding occurs so that teams are secondarily ranked by their percentage.) And then that team is tipped to win the following week, but it remains on 15 wins, now "15 (15.3)". What has happened is the number of probable wins hasn't risen by enough to be rounded to a higher number: it has earned 0.6 more probable wins but this still rounds off to 15. The predictor is saying it's still most likely this team will be on 15 wins, after accounting for the likelihood that some of its tips will be wrong.

 

[The Swans Blog]

PILGRIM hasn't been good enough for long enough. Last year it did 72.0%, the same as ISTATE, with slightly better margin accuracy, but it's done a little worse the three years before that.

That's interesting because it's already 5 up on tips this season compared to Ol'Faithful. Is there any reason for that? You said it's more sensitive to recent form, but would it have picked Lions last night? Or Eagles the week before?

 

[TBellicious]

How accurate has the ladder predictor been in the first 10 rounds of a season in comparison to the final standings over past years?

 

[Final Siren]

That's interesting because it's already 5 up on tips this season compared to Ol'Faithful. Is there any reason for that? You said it's more sensitive to recent form, but would it have picked Lions last night? Or Eagles the week before?

Most likely, the reason is simply that if you have 157 people roll dice, you'll get a few who roll lots of 6s. But once you notice that, you don't find those people more likely to keep rolling 6s than anyone else.

 

[Final Siren]

How accurate has the ladder predictor been in the first 10 rounds of a season in comparison to the final standings over past years?

I'd have to crunch some numbers to get a good read on that. Maybe I could compare it to the actual ladder, round by round, for some frame of reference. But my guess is it's somewhere in the realm of a very good human predictor: better than most, not better than everyone.

One great way to predict ladders, though, is to ask everyone and average their answers. In 2012 I did this thread and in 2014 GOAUS! did this one where we made consolidated BigFooty ladders based on everyone's individual pre-season predictions.

The result was not only pretty good, but also quite a bit better than the average ladder used to create it! This is a "wisdom of crowds" effect where people can make smarter decisions as a group than as individuals, because you average away all the crazy outliers. In 2014, the average BigFooty poster's ladder was out by 40 rungs, but the consolidated version, created by mashing them all together, was only out by 28 rungs. Like magic!

 

[Richo83]

May not be a popular question but;

Does the squiggle overrate thrashings of poor teams?

The only teams that West Coast have beaten that are worth anything are Port and GWS - GWS who are hard to read (young team travelling across the country) and Port (having a miserable season).

This week West Coast have the Saints - could get nasty again.

Eh, if beating sides by massive margins was easy, then it would happen any week. Good sides tend to beat up on weak sides by a higher degree and more often than mediocre sides, so there is something into reading into a 100 point win against Melbourne. I think the problem with West Coast is less that we read too much into their wins against poor teams and more about the fact that it's early days and a lot can still change.

 

[Randomizor]

I keep thinking that North are top 4 material.
The Squiggle just has us tooling around in the zone of mediocrity.

Are they though?

I guess with Port looking unlikely to make it you have as good of a chance as any of the other 9 teams on the same points as you or more

 

[TBellicious]

I'd have to crunch some numbers to get a good read on that. Maybe I could compare it to the actual ladder, round by round, for some frame of reference. But my guess is it's somewhere in the realm of a very good human predictor: better than most, not better than everyone.

One great way to predict ladders, though, is to ask everyone and average their answers. In 2012 I did this thread and in 2014 GOAUS! did this one where we made consolidated BigFooty ladders based on everyone's individual pre-season predictions.

The result was not only pretty good, but also quite a bit better than the average ladder used to create it! This is a "wisdom of crowds" effect where people can make smarter decisions as a group than as individuals, because you average away all the crazy outliers. In 2014, the average BigFooty poster's ladder was out by 40 rungs, but the consolidated version, created by mashing them all together, was only out by 28 rungs. Like magic!

Also i would like to know what total tips the squiggle is on this year?