The Forest from the Trees: A post for the stats nerds


#161

Have you seen advertising plastered all over it?

Ha.


#162

AFL tables is a guy called Paul haha, not AFL affiliated. I am starting to believe everyone scrapes. I am just not sure on the legality of using the data for profit?


#163

Sad and sorry situation, only 6 of our players playing well above average against the bottom team, according to Champion Data.


Caveat: This is Champion Data and is less than perfect as we all know.

Well above average players: Heppel, Zaharakis, Ambrose, Hooker, Fantasia, McKenna.

Well below average Players: Merret, Hurley, Myers, Stewart, Goddard, Tippa, Stringer, Parish, Laverde

Parish, Tippa and Goddard need to be rested. But realistically, we are not going to omit more than 4 players, and look at the ones who are not improving week on week

They need to start with Laverde who was poor in the previous week, as were Goddard and Parish ( need a rest) send Stringer, down to the VFL


#164

This is really excellent


#165

Thanks for posting that mate.

They have some really interesting perspectives.


#166

Champion data is part owned by the AFL and has AFL people like Dillon as Directors. I don’t know whether clubs have alternative sources or whether reliance on Champion Data is imposed on them for data analysis ( and whether they have to pay for it).


#167

Apparently Dangerfield played well, but Selwood and Ablett were poor. Ablett in particular does not like being caught and crunched. Snuff those guys out and Geelong are not that good.


Caveat: This is Champion Data and is less than perfect as we all know.

Well above Average: Merret, Hooker, McKenna, Tippa, Stringer, Langford , Baguley, Laverde
Above Average: Saad, Fantasia
Well Below Average: Ambrose, Bellchambers, Goddard. ( Goddard played better than the stats showed )

Consistent across 4 quarters. Zaka, Ridley, Smith
Big last quarter: Stringer. Dea , McKenna.
Drop offs in last quarter ( tired or sore?) Tippa, Fanta , McKernan

Rising Star: Jordan Ridley. As good a debut as you would see,


#168

Hawkins beat Hooker?

Ok


#169

Absolutely not, when the game was there to be won.

Did we switch off in the last quarter , or were some key players gassed, ?


#170

One of the best things about these stats is the quarter by quarter break down. Hawkins generated most of his points in the 4th where he DID kick 3 goals. Hookers intercept marking dropped off then, no doubt because the fast Geelong repeat entries didn’t allow him the chance to set himself.


#171

Hooker was best on by a mile.

Just surprised to see Hawkins with the higher rating.


#172

Champion data from the outside appear to be a big average analytics provider. But as a source of - collection of - data for stats I think they have a monopoly on almost all sports in Australia.

I would hope that the club takes what it can get from them, mostly to know what other clubs get and save on the basics. But are employing some smart people to value add on top and push the analytical boundaries. I doubt it though.

Any club that isn’t getting into machine/deep learning in a big way to analyse recruiting, individual and team performance is already behind. This is something that should be very closely guarded internally.


#173

I think Aussie rules is a LONG way behind on this stuff.

Port Adelaide hired the blogger from figuringfooty.com in the off season, which indicates to me clubs are looking for analysts who use data in insightful ways.

Hpnfooty do brilliant draft analysis each year, and some of the principles they’ve uncovered I reckon Hawthorn have been taking advantage of for years.


#174

Goals give you a lot of points. But junk time goals when the games over should be even less.


#175

One of the Essendon analysts was on Grandstand recently


#176

Thanks for the tip.

Just found the podcast, but yet to have a listen:


#177

What a load of crap these stats are. We flogged em well and truly, not by a little bit or a 20% influence, we totally embarassed them and their supporters knew it too. We were ten times better.


#178

90+ tackles and 14 kms of 2 way running is a lot of hard work
No doubt we switched off intensity in the last. I suppose if you know you have the 4 points, thoughts surely change to recovery and pulling up in good nick ready to reset for the next game.


#179

Whether we turned off or not, Champion Data’s algorithms are clearly screwed when acts that make no difference to the result can be rewarded so much.

They claim that their stars try and measure influence on the game, and then serve up these numbers that somehow give Geelong 4 of the 6 most influential players on the ground in VA match that they were never even vaguely competitive in.

Those are numbers that would make me throw out my algorithms and start again.


#180

CD’s raw data is very good. But those player ratings are straight up garbage. They’re just preposterous. Utter rubbish.