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


#181

The raw data is for club use.
The ratings are just for forums.


#182

Fxd


#183

This is correct I watched the scoreboard at the game and the game was probably over by quarter time, its just that most Essendon fans would not have made that call until half way through the last quarter

One common criticism is that the ratings are bs. I agree they not of themselves perfect in any way shape or form. However, they can be useful to compare players performance by their own benchmarks, viewed through the prism of the ratings.

Examples ( that we dont need to know because we saw it for ourselves.)
In the above ratings Ablett was rated as 25% of his own average rating. I reckon we all agree Ablett was barely AFL level in that game. He was so bad he was looking for unnecessary 1-2 s to try and pump up his stats.
Selwood was 40% of his average rating. We would all agree he had a lot of useless possessions.

The presentation shows stats in each quarter. Which may not be immediately obvious

Examples Conor was quiet first and third quarters. Stringer did nothing in our big first quarter, but lit up in the last. Tippa was invisible in the last . Matty Dea had a big last quarter, and Hooker fell away.

I don’t think anyone can deny Hawkins got off the leash in the last. But Dangerfield BOG? Give me a break. He accumulated in his usual way. That is because Champion Data regard Dangerfield as their prototype best AFL player, and the algorithm is tuned to rate him as the best player. They admitted that last year.


#184

sigh the player ratings are built around effective disposals, and are apparently the most advanced system CD has for measuring that. It not only measures whether the touch ‘hit a target’, but also the angles and distance of the kicks. Dangerfield played a good offensive game (lots of kicks, 2 goals). By their own admission it does not measure negating actions. This came up on twitter one time when Tom McDonald (I think?) ranked best on ground despite having 6 goals kicked on him. It was measuring what he did, not what he stopped.

Use a statistical system for its intended purpose.


#185

He’s done that his entire career.


#186

One critical stat that is not publicly available (and will always be difficult to measure in AFL as we don’t play many matches in a season) is wins above replacement (WAR). Google it if you need to, but it attempts to measure a players role and how much better they are than the person that would do the same role. This acknowledges that certain roles (i.e. key play making midfielders, loose defenders) are statistically ‘privileged’.

This is why you get people making comments like “I’d get 30 touches if I was playing Hurley’s role.”

In effect, Ridley replaced Hurley and got 25 possies, meaning as a replacement he was pretty good, and if Ridley maintains that (pretty unlikely given he’s a first year player) then Hurley’s value to the team and his impact in general has been massively overrated.

This WAR idea was always why I was sceptical about Ablett at Geelong. Both he and Dangerfield have built careers out of being the play maker, they don’t play a team first role in the way that Selwood does, and I’m not sure the value add is what you’d expect. This is probably why the Cats have been trialing them both up forward for extended stints to find a position where the “value add” is higher.


#187

Stats don’t show that Ablett hates being hammered and we hammered him. Zerret ground his head into the turf big time and Ablett was dazed after that. 17 disposals. 1 mark, 1 tackle, 1 inside 50, 1 clearance. The worst game he has played in 5 years.

image

In fact we hammered all their midfielders and they were not prepared to be ambushed in that way.
Ablett, Duncan, Menegola, Selwood, all below their own average *.

The quirk in the CD system is that because Dangerfield is the best player in the game*, he only had to play an average game for him and he was rated best on ground*

*According to Champion Data.


#188

I agree WAR value are a better metric for measuring a players impact. But I’m not sure it’s been tested effectively at AFL yet?

I haven’t seen it being used extensively.

I wonder if the variance in how a particular player plays a position and the role of positions differing in teams make it much more inaccurate?

What I find funny about CD is they are a full time stats organisation and have the insights of somebody who’s done a cert II in excel. They either don’t realease much publically or are all about collection that analysis.


#189

Surely the stats we get are a subset. For instance the most simplistic analysis we get publicly for disposals that everyone goes wow about is disposals.

Illustrating this is the different form of the stats in the VFL.
Marks and handball receives are given for posessions while on the disposal side its kicks effective, kicks ineffective and handballs effective and ineffective.

Given the VFL data its possible to differentiate between a very effective handballer (Clarke) and a very effective kick (Langford ) Its just not possible to do that with the AFL stats we receive.


#190

I think we get about 5% of what available, as if you get GPS, offensive and defensive metrics with a whole bunch of meta data about those parameters then you could almost create the entire game virtually.

Which is why I suggest CD are poor in the analysis and focus on collection. As they could be doing some substantially more significant things with the level of data they appear to have.


#191

I reckon that’s valid. AFL is one of, if not the most data rich sport in the world and yet a lot of what we see as the public is just so obviously flawed.


#192

There is a bit of obsession with hiding things in this sport.
So what you get, is the rubbish ‘commentators’ trot out with their reverse-analysis and crap terminology like ‘West Coast Web’. A moderately intelligent monkey can reverse a section of play and then go on to explain how it was all planned. In monkey-language, of course.


#193

That’s less about hiding things as it is about catering to the majority of viewers level of intelligence. I’m sure you’ve seen the nuffies page on facebook…


#194

I don’t use fb man hahahaha
And it’s not because I’m a Luddite.


#195

get your daily dose of nuffies around here?


#196

I try not to, but yes :slight_smile:


#197

I think it’s more about maintaining the value of their service. If they are shown up based on a cheap easy access to data then they are nothing but data entry monkeys which isn’t valuable. By holding it back and trying to sell it as value add they are going for a higher $. Which is very AFL.


#198

If only we could get to hear these monkey commentators - it would be a big relief from the current farkwits inflicted on us.


#199

I think Champion Data send the data to David King to be processed into it’s highest use.


#200

… OK