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


#201

Thanks for the details BBB, I hadn’t spent any time investigating the details of these ratings, because who has time…and because I’d decided that it wasn’t worth my time given they were clearly flawed.

Of course CD know what it does and doesn’t measure, but the choice to name it “AFL player ratings” and put it out there every week is deliberately misleading and detrimental to both their brand and the wider field of sports statistics.

A key component of any statistical analysis is the interpretation and presentation of results. CD could quite easily name these “ratings” as “offensive efficiency ratings” or some other name that provided additional detail and context to the reader; however, they have made a decision (likely commercially driven) to present them in the way they they do.

I think that they have it all wrong in CD world. (And AFL data world). The algorithms are what makes CD anything more than a data entry company. If they believe themselves to actually be an analytics company they should be providing better analysis for public consumption. (given that they are choosing to provide analysis for public consumption).

No one is asking them to publish their algorithms, and tbh I don’t think we even want their most highly valued metrics. But we want something more than the Herald sun level statistical analysis that we currently get.

I also think what we want (as a minimum) is historical raw data sets released for public research and analysis. I’m sure we’d all happily pay a fee if the price was right. It wouldn’t be as interesting as weekly data dumps, but I accept that we’re not getting it because the AFL have enforced an anti-conpetitive monopoly situation that they won’t want to break.

Imagine the interesting information that would come out (and fill summer news pages) if they dumped the season’s complete raw data package to the community on October 15.


#202

Frosty,
There was a change this year. Player ratings has been downgraded.
A large subset of the raw data is being made available.
Its on the AFL web site under stats and its called Stats Pro.

Example.

Stats like contested defence loss percentage best. Hooker, Saad and Ambrose .
Defence half pressure acts best . Merret Smith, Clarke.
Forward 50 Ground ball gets best . Green String Smith.
Spoils best Ambrose , Hartley, Brown


#203

Stats Pro is still going to be controversial, even though they are raw stats, For instance Hurls and Hepp are often criticised for poor kicking.
But these stats have them both in the top 4.
Ridley is a small sample, but believable because his disposal is " elite" ( for his age)

That Effective Kicks stat is part qualitative and subject to error.


#204

The raw numbers still need to be considered with some context. Huge improvement in the stats the AFL are making available now though.


#205

Just perusing some of the defensive numbers on Statspro…
And assuming I’m interpreting them correctly, Ambrose is the equal of Alex Rance as a key defender this year.


#206

Of course I’m absconding with the urine there… that only applies specifically to one on one defensive ability.

Rance is clearly better offensively than Ambrose.

Hooker is clearly better than both. Interestingly, Lever is proving himself more adept one on one than he has in the past.


#207

It is interesting to compare the hotly disputed positional selections using this tool.
For example
Hartley V Ambrose, where Ambrose is shading Hartley but not by a huge margin overall, key differences in Ambrose favour being Tackles, Spoils,Rebound 50s, Inside 50s and Contested possessions. So his selection may be more about bodywork, mobility and endurance than other stats.


#208

Just had a look at “statspro”

Yeah, nah, that’s not what we looking for, unless there’s a simple way to scrape the entire set.


#209

If you want to get the raw dataset, you will have to pay $$$

No doubt they sold this Stats Pro product to the AFL to suck some cash flow out of the massive AFL revenue stream in a "legitimate " way.


#210

CD is at least partially owned by the AFL


#211

And this was my point.

There’s no inherent value in knowing whether player X provides more inside 50s than player Y as a statistical measure.

That is something that a person watching a game can pick up (and if you really wanted to you could go and watch every match and generate all that data by yourself)

The value of sports stats is in getting all of this raw data together and using some combination of algorithms to derive insights that are beyond what you can see with your own eyes.

Stats pro just presents what you can already observe in an easy(ish) to digest manner. But it isn’t sports analytics.


#212

Actually I would like to enjoy watching footy. Collecting stats is too much like hard work.


#213

I am almost certain there is no way to buy the stats pro data in one package. I also know/think they don’t have historical players, even though you can track current players with the new set of metrics back to 2012. When you are doing whole game analysis, this is a massive flaw.

Scraping the data is the only way that I know of. Would love it if anyone can 1. find an alternate way to get it or 2. find the historical data so I can scrape the entire set. If someone does that I will post it to whoever wants it.


#214

#215

Play Daniher Stewart Smack Hurley AND HARTLEY at half-forward.

Myers at full-forward; if he gets the ball within thirty metres he’s a legit chance.


#216

I mean, have we ever tried Myers one out in the goal square? Give the kid a chance Woosh


#217

GWS Game CD stuff.

Heath Shaw probably saved GWS from a 10 goal loss,
LOL. Jeremy Cameron rated 2nd worst player on the ground. Ambo cancelled him out.

Stringer consistently showing he has the tank to lift in the last quarter. ( pregame Chicken Fillet Burgers work)


#218

McGrath went under the radar based on those rankings.

He doesn’t stand out that much because he rarely makes any mistakes.

He is tracking very well.


#219

Which is ironic… given the lazy downhill skiing ■■■■ personifies the issues with GWS… Played most of the match as the spare man in defence and didn’t register a single ■■■■■■■ tackle…

He was insipid. Another reason why CD ratings can eat a whole bag of dicks.


#220

Matt Guelfi’s impact in the last quarter was far more pronounced than that chart suggests also.