Certified Legendary Thread Race for the flag, in squiggly lines

Suma Magic · Mar 15, 2016

cryptor said:
You'd do even better on top of this if you don't blindly follow it but adjust for injuries, etc.

I find that my instincts and analysis usually backfire.

JaxsonB · Mar 16, 2016

As someone that loves maths and statsing (while only being at a first year uni level for it, so forgive any oversights ^^'), I did some more sleuthing about Squiggle's accuracy about a certain thing: games that it predicts with confidence.

First point - I have to decide what I consider to be "Squiggle confidence". I went with >21.5 points (3.5 goals). I feel that it's low enough that it doesn't render any result meaningless (of course the sheer majority of games predicted to be won by 40+ points are going to result in correct tips), but high enough that there's at least some confidence in the tip. If you're not happy with 21.5 you can balloon it out to whatever you think is fair, but that's what I went with for this.

In any case, what I did was look at Squiggle's tipping accuracy *only* for games that it felt confident in, using the point above (team tipped to win by 22 or more). Given that I had to do this for 23 rounds in each season, I only went back as far as 2013, but this is what I found:

In 2015, Squiggle was accurate 77.9% of the time when tipping a team to win by 22 or more points.
In 2014, Squiggle was accurate 80.8% of the time when tipping a team to win by 22 or more points.
In 2013, Squiggle was accurate 83.8% of the time when tipping a team to win by 22 or more points.

The thing to keep in mind with these results, however, is that I haven't recorded how confident Squiggle was in each loss. It's likely that the losses fell closer to the 22 point minimum than they did to, say, a 70 point tip, so to say that *any* tip that is at 22 points or more is about a 4 in 5 chance of being accurate is at least somewhat fallacious.

However, it's common sense that Squiggle performs better as the season goes on and formlines start to become apparent. Looking at the same figures, except only from rounds 11 to 23:

In 2015, Squiggle was accurate 82.8% of the time.
In 2014, Squiggle was accurate 83.8% of the time.
In 2013, Squiggle was accurate 86.0% of the time.

Let's be real - those are some amazing odds. There's still the same issue as above (in all likelihood the incorrect calls it made were likely closer to the 22 lower limit, as opposed to being a large margin), but it's at least interesting to see.

While I believe many human tippers would get similar rates of accuracy, it's interesting to see how accurate Squiggle can be by following a simple formula.

Randomizor · Mar 16, 2016

JaxsonB said:
As someone that loves maths and statsing (while only being at a first year uni level for it, so forgive any oversights ^^'), I did some more sleuthing about Squiggle's accuracy about a certain thing: games that it predicts with confidence.

First point - I have to decide what I consider to be "Squiggle confidence". I went with >21.5 points (3.5 goals). I feel that it's low enough that it doesn't render any result meaningless (of course the sheer majority of games predicted to be won by 40+ points are going to result in correct tips), but high enough that there's at least some confidence in the tip. If you're not happy with 21.5 you can balloon it out to whatever you think is fair, but that's what I went with for this.

In any case, what I did was look at Squiggle's tipping accuracy *only* for games that it felt confident in, using the point above (team tipped to win by 22 or more). Given that I had to do this for 23 rounds in each season, I only went back as far as 2013, but this is what I found:

In 2015, Squiggle was accurate 77.9% of the time when tipping a team to win by 22 or more points.
In 2014, Squiggle was accurate 80.8% of the time when tipping a team to win by 22 or more points.
In 2013, Squiggle was accurate 83.8% of the time when tipping a team to win by 22 or more points.

The thing to keep in mind with these results, however, is that I haven't recorded how confident Squiggle was in each loss. It's likely that the losses fell closer to the 22 point minimum than they did to, say, a 70 point tip, so to say that *any* tip that is at 22 points or more is about a 4 in 5 chance of being accurate is at least somewhat fallacious.

However, it's common sense that Squiggle performs better as the season goes on and formlines start to become apparent. Looking at the same figures, except only from rounds 11 to 23:

In 2015, Squiggle was accurate 82.8% of the time.
In 2014, Squiggle was accurate 83.8% of the time.
In 2013, Squiggle was accurate 86.0% of the time.

Let's be real - those are some amazing odds. There's still the same issue as above (in all likelihood the incorrect calls it made were likely closer to the 22 lower limit, as opposed to being a large margin), but it's at least interesting to see.

While I believe many human tippers would get similar rates of accuracy, it's interesting to see how accurate Squiggle can be by following a simple formula.

Be interesting to see what the accuracy would have been if you simply always tipped the favourites.

wahawk · Mar 16, 2016

Randomizor said:
Be interesting to see what the accuracy would have been if you simply always tipped the favourites.

I did that last year in my tipping.
I won the comp I was in but no where near winning some others

iluvparis · Mar 17, 2016

wahawk said:
I did that last year in my tipping.
I won the comp I was in but no where near winning some others

Generally tipping the favourites each week is good enough to finish top 10-15% in most tipping comps but you'll pretty much never win.

Final Siren · Mar 22, 2016

Footy only days away! It feels like teams could move a lot this year. I mean, it always does, but ESPECIALLY this year. I'm looking forward to finding out about:

Hawthorn: Last year they delivered one of the more emphatic premierships of the last two decades. They actually look about as formidable as any team could be. Can they keep it up?

Fremantle: Squiggle says they were ordinary in the second half of 2015 and expects them to fall like a stone this year.

Port Adelaide: Which was the aberration, 2014 or 2015?

Geelong: Hell of an off-season, which the squiggle knows nothing about, so are the Cats about to launch?

Carlton & Essendon: How bad, exactly? GWS 2013 bad? Because that was pretty bad.

And a heap of teams have reasonably serious buzz. You wouldn't be surprised to see a charge from Collingwood, the Bulldogs, GWS, Adelaide, or even Melbourne, after an impressive pre-season. And you'd expect West Coast, North Melbourne, and Sydney to all be up there.

Could be a very squiggly year!

smeltitdeltit · Mar 22, 2016

JaxsonB said:
As someone that loves maths and statsing (while only being at a first year uni level for it, so forgive any oversights ^^'), I did some more sleuthing about Squiggle's accuracy about a certain thing: games that it predicts with confidence.

First point - I have to decide what I consider to be "Squiggle confidence". I went with >21.5 points (3.5 goals). I feel that it's low enough that it doesn't render any result meaningless (of course the sheer majority of games predicted to be won by 40+ points are going to result in correct tips), but high enough that there's at least some confidence in the tip. If you're not happy with 21.5 you can balloon it out to whatever you think is fair, but that's what I went with for this.

In any case, what I did was look at Squiggle's tipping accuracy *only* for games that it felt confident in, using the point above (team tipped to win by 22 or more). Given that I had to do this for 23 rounds in each season, I only went back as far as 2013, but this is what I found:

In 2015, Squiggle was accurate 77.9% of the time when tipping a team to win by 22 or more points.
In 2014, Squiggle was accurate 80.8% of the time when tipping a team to win by 22 or more points.
In 2013, Squiggle was accurate 83.8% of the time when tipping a team to win by 22 or more points.

The thing to keep in mind with these results, however, is that I haven't recorded how confident Squiggle was in each loss. It's likely that the losses fell closer to the 22 point minimum than they did to, say, a 70 point tip, so to say that *any* tip that is at 22 points or more is about a 4 in 5 chance of being accurate is at least somewhat fallacious.

However, it's common sense that Squiggle performs better as the season goes on and formlines start to become apparent. Looking at the same figures, except only from rounds 11 to 23:

In 2015, Squiggle was accurate 82.8% of the time.
In 2014, Squiggle was accurate 83.8% of the time.
In 2013, Squiggle was accurate 86.0% of the time.

Let's be real - those are some amazing odds. There's still the same issue as above (in all likelihood the incorrect calls it made were likely closer to the 22 lower limit, as opposed to being a large margin), but it's at least interesting to see.

While I believe many human tippers would get similar rates of accuracy, it's interesting to see how accurate Squiggle can be by following a simple formula.

sorry, but are those results based on 22 points meaning that when squiggle picks a margin of 22 or more, the margin is generally 22 or greater, or is when the squiggle tips 22 or more, thqt team wins (regardless of margin) approx 85% of the time?

iluvparis · Mar 23, 2016

smeltitdeltit said:
sorry, but are those results based on 22 points meaning that when squiggle picks a margin of 22 or more, the margin is generally 22 or greater, or is when the squiggle tips 22 or more, thqt team wins (regardless of margin) approx 85% of the time?

Latter

run2dmax · Mar 23, 2016

Trying to move some teams around and run a prediction, but no prediction is loading.

Rotayjay · Mar 23, 2016

Perhaps we should have a new Squiggle thread for each season.

Randomizor · Mar 23, 2016

Looking forward to the squiggle not understanding Freo again this year

Suma Magic · Mar 24, 2016

Looking forward to Freo fans not understanding the squiggle again this season.

iluvparis · Mar 24, 2016

Randomizor said:
Looking forward to the squiggle not understanding Freo again this year

Looking forward to Freo fans endlessly downplaying the factual/statistical nature of the squiggle as their 'vibe' about their team is more informative.

Randomizor · Mar 24, 2016

When the premiership flagpole shows Port closer to a flag than Freo, when they didn't even make finals, you know someone has stuffed up some maths

iluvparis · Mar 24, 2016

Randomizor said:
When the premiership flagpole shows Port closer to a flag than Freo, when they didn't even make finals, you know someone has stuffed up some maths

I don't think you understand probability very well.

Mad_Hatter · Mar 24, 2016

Suma Magic said:
Looking forward to Freo fans not understanding the squiggle again this season.

Randomizor said:
When the premiership flagpole shows Port closer to a flag than Freo, when they didn't even make finals, you know someone has stuffed up some maths

That was quick.

Gydafud · Mar 24, 2016

Looking forward to seeing some movement on the squiggle after tonight!

Illinois Nazi · Mar 24, 2016

Rotayjay said:
Perhaps we should have a new Squiggle thread for each season.

Yes please!

Final Siren · Mar 24, 2016

run2dmax said:
Trying to move some teams around and run a prediction, but no prediction is loading.

Ah, thanks for letting me know. Now working again.

Randomizor · Mar 24, 2016

iluvparis said:
I don't think you understand probability very well.

Port was more likely in 2015 to win the premiership? 9th placed Port?

dcutter · Mar 24, 2016

Randomizor said:
When the premiership flagpole shows Port closer to a flag than Freo, when they didn't even make finals, you know someone has stuffed up some maths

Not really. The squiggle offers statistical insight as to why no one rates Freo and why you always fail when it counts.

I rate port a hell of a lot more than Freo, I think most do.

Randomizor · Mar 24, 2016

dcutter said:
Not really. The squiggle offers statistical insight as to why no one rates Freo and why you always fail when it counts.

I rate port a hell of a lot more than Freo, I think most do.

More that Port were 9th, so was impossible for them to win it all, yet the flagpole showed them as more likely than Freo. Doesn't make sense. Unless it's not an each year thing, but on a rolling basis.

That's nice. It didn't think we'd make minor premiers and we did. So it was completely and utterly wrong there. Statistics be damned.

When does it "count" by the way? If the premiership is the only yardstick then we have failed and have 16 other teams that failed along with us... Pretty stupid statement.

rippersnipper · Mar 24, 2016

Randomizor said:
More that Port were 9th, so was impossible for them to win it all, yet the flagpole showed them as more likely than Freo. Doesn't make sense. Unless it's not an each year thing, but on a rolling basis.

That's nice. It didn't think we'd make minor premiers and we did. So it was completely and utterly wrong there. Statistics be damned.

When does it "count" by the way? If the premiership is the only yardstick then we have failed and have 16 other teams that failed along with us... Pretty stupid statement.

Someone correct me if i'm wrong but it is a rolling thing which is then applied to that particular season. Hence port's flagpole was higher due to their previous two seasons (their huge increase in performance) as opposed to yours where you made a grand final and then were knocked out in a semi final. This would explain why their's was higher. Of course if this is the case then your's should be higher than port's this year as you perfprmed better last year. I'm not sure what the weighting is though. Or even if i'm correct. It would make sense if this is what the squiggle does though.

iluvparis · Mar 24, 2016

Randomizor said:
More that Port were 9th, so was impossible for them to win it all, yet the flagpole showed them as more likely than Freo. Doesn't make sense. Unless it's not an each year thing, but on a rolling basis.

That's nice. It didn't think we'd make minor premiers and we did. So it was completely and utterly wrong there. Statistics be damned.

When does it "count" by the way? If the premiership is the only yardstick then we have failed and have 16 other teams that failed along with us... Pretty stupid statement.

But it was correct in showing despite being on top of the ladder you weren't a genuine threat at the flag.

Thoughts and sympathies with you at this difficult time.

Randomizor · Mar 24, 2016

iluvparis said:
But it was correct in showing despite being on top of the ladder you weren't a genuine threat at the flag.

Thoughts and sympathies with you at this difficult time.

Somehow think that you're going to have a worse year watching footy than I am.

Log In or Create New Account

Certified Legendary Thread Race for the flag, in squiggly lines

Brownlow Medallist

Team Captain

Brownlow Medallist

Norm Smith Medallist

Import Whisperer

Mr Squiggle

Premiership Player

Import Whisperer

Senior List

Brownlow Medallist

Brownlow Medallist

Brownlow Medallist

Import Whisperer

Brownlow Medallist

Import Whisperer

Brownlow Medallist

Premium Platinum

Brownlow Medallist

Mr Squiggle

Brownlow Medallist

Club Legend

Brownlow Medallist

What up son, whatever

Import Whisperer

Brownlow Medallist

Certified Legendary Thread Race for the flag, in squiggly lines

Similar threads