I think the problem with the Graph is there is no historical reoreseantation. What’s the premiers squiggle normal distribution, top 4, your team history (last year, last 3, 5 etc).
It’s hard to make any kind of assessment with nothing to compare it to.
It’s a real nice thing to play around with, but it’s still early days in the season to mean a lot. Maybe mid-way it will have more use.
Its based on a couple of years of data - its a simple statistical model. Sure the variance in R1 is highest than any other point in the season, but I don’t get your beef with it?
Dont take it personally Trevor, Deckham hates stats at the moment.
To be fair - this is an odd choice of thread for me to hang out in 
You are a ■■■■■■■■■
I’m pretty sure Decks has a beef with any model.
You love it, you just have realised that yet!!
That squiggle thingy is pretty sexy.
Squiggle rate GWS as the best defenders in the comp at about 67, while we are rated as the 4th best forward line at 57. So, we are well matched at that end of the ground. Our defence is a bout mid table, which is a match for their relatively poor attack similarly at about midtable. Very hard to pick this result on the basis of that simplistic comparison.
But the models favor GWS to win at 55-60%. Mind you, the models are often wrong, so don’t get too worried.
“All models are wrong. Some are useful”-George Box
One of the highly-ranked models last year got almost all of our tips wrong last year (only started getting it “right” around round 15).
To be fair, we did win all the ones we shouldn’t have (and vice versa) in that period…
For those that don’t understand why I might not rate Cripps as high as the rest of you this article might explain why.
As I have highlighted below, two-thirds of all scores are actually created from turnovers.
It doesn’t take a genius to work out that a player that does most of his best work at the coal face is not going to have a huge influence on the game if most of the scoring comes from turnovers.
Anyway that’s just my opinion.
An AFL side can win possession of the ball from one of three basic sources: clearances, kick-ins after behinds, or turnovers.
Although many pundits often focus on the former, it is the latter from which scoring is most often generated. In fact, about two-thirds of all points usually come from turnovers.
Richmond’s game plan exploits this key fact of modern football.
The number of turnovers per AFL game has increased by 13 per cent over the past four seasons, from 126 per match in 2015 to an average of 143 this season.
Games involving the Tigers feature even more: 155 per game this season, up from 125 four years ago. Crucially, no side is better at forcing its rivals into mistakes whilst minimising its own.
On average, the Tigers give up the ball nine times fewer per match than their opponents.
“Our game is a turnover game,” said Hardwick at Wednesday’s pre-match media conference.
“We acknowledge that, and most sides know that, so the way we set up [against Hawthorn tonight] will certainly be turnover-based.”
I think you meant to say you don’t rate "the importance of " Cripps to a winning game plan .
As a contested clearance beast, he is elite. But I guess your article goes some way to explaining the fact that FCarlton finished last, even with one of the top 5 clearance mids in the comp playing for them.
Compare the Cripps game 2 weeks ago to the Selwood game a week ago. 38 disposals, starting on a wing and ranging through the midfield picking targets and executing very very well. Its not something Cripps is particularly good at. Selwood can do both. Hepp can play on the inside or outside. They are more complete players.
I stated clearly that Cripps was elite as a clearance player but his overall impact doesn’t mean he is an elite player.
What those numbers do highlight though is that a high clearance team like Collingwood can hurt Richmond but a lot has to go right to do so.
As it played out Collingwood got them in the prelim.
So 82-73 is the greatest differential in the league. They probably score about as much from that as we do from having the best midfield combo…
EDIT: a quick search indicates the better teams may be around 1 point per turnover. I also saw that Hawthorn in 2015 had a turnover differential around 23…
I made a predictive model if any of you nerds are interested. Just a bit of messing around, but it does okay. Not sure how to make the metadata on my links not look ■■■■■■ though.
https://www.thefatstats.com/fatstatsblog/autotippa-ii
https://fatstats.shinyapps.io/Auto_Tippa_II/
You’ve made the Squiggle!
Alas AutoTippa doesn’t like our chances on the weekend, nor do most other models. Only two, matter of stats and afl gains, like our chances.
With that said, the margins are close and most models factor in home ground advantage. I suspect if we were home it would be reversed.
AutoTippa apparently takes into account personal? I’m surprised the Giants outs didn’t have more of an effect on its predictions . (Only skimmed your model mind)
Fantastic. Looking forward to reading more when I get time.
Interested in the variable selection used to trim the variables used down for the player influence. I assume you used an automated process, did it throw out variables that you thought would be critically important or vice versa. And did you end up tweaking manually at all?
Forest for the trees indeed!
