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

Football Discussion Dustin Fletcher Hangar
762

image

How the fark do they get some of these? Collingwood have Grundy (1) Treloar (11) Pendlebury (26) Adams (54) Sidebottom (69) & Crisp (92) thats ranked by average SC points based I thought on CD’s figures. North have Goldstein (12) Cunnington (34) Higgins (65) & Ziebel (84) so how have they ranked North higher than Collingwood let alone as second best midfield?

Some of it makes sense & some is just so wrong its not funny. How can Geelong have the 16th ranked defence - they conceded the least points for. Even if you ignore results & look purely at stats they still have Stewart (8), Blicavs (30) Guthrie (61) Tuohy (90) Taylor (91) & Henderson (99) compared to Freo - Ryan (13) & Blakely (64) are the only dockers inside the top 100 average points. How can Freo be ranked higher?

763

Without getting into too much detail, most statistical outcomes can be radically altered by varying two things.

Your initial decisions about what to include or exclude from your categories. In this case List of Club defenders for example. Who do you call a defender at any Club. (CD is very dodgy on that)

Any weighting’s you may apply to factors within your overall sample. In the case of CD & footy they make assumptions about the importance of things like “winning clearances” for example that are often at odds with reality.

Remember stats are just a model of the real world.
Some models are more accurate and reproducible than others.
Some models are honest enough to admit both their possible margin of error, and their past success at modelling the real world accurately.

CD is notable in that it refuses to publicly do any of the above.

5 Likes
764

They hedge their bets. Everything has a probability tag. So in a universe somewhere, Essendon will be minor premiers in 2020.

I circled the chance we will finish on top of the ladder, according to Squiggle. Around 1 in 100 as best I can see.

image
image905×154 57.3 KB

765

I agree, they should start showing some sensitivity analyses.

766

Well, assuming those percentages are correct, there’s a 33% chance that we’ll be 2-3 after 5 rounds. So it’s actually more likely we won’t be 2-3 going into Anzac Day.

The odds of our number of wins from 0-5 based on those percentages looks like

image
image1052×61 276 Bytes

2 Likes
767

Worth looking at the video in this article that credits an inefficient clearance to Clarke. Gives a great illustration of why efficiency stats in particular are a very dubious reflection of what we might intuitively think of as “efficient”.

3 Likes
768

To be fair to CD, this is one area where the definition is clear. Clearance means it gets out of the immediate area in your team’s hands.
A lot of people think a clearance is just who gets first hands to it.

TBH to me that clip just looks like an otherwise good player rattled, expecting pressure, & making a bad decision.
Nothing to do with game plan, or even any of the 3-4 players involved.
If DP was more aware, he stays on the move,
draws the nearest Weagle towards him, then released McGrath inside.

Eagles set up for that 3 on 2, with McGovern or Hurn as the 3rd up.

769

Sure, but the article notes it’s credited as an ineffective clearance to Clarke. I would have thought either it’s Parish’s clearance and it’s ineffective or it’s Clarke’s and it’s effective. Crediting it to Clarke and then saying he did it ineffectively because of Parish’s subsequent kick is not the obvious interpretation to me.

2 Likes
770

No.
First disposal in a chain when your team clear the area with it.

771

Which was kind of my original point, badly made I guess. The definition is clear, and it applies to the piece of play. But it’s not necessarily intuitive to most people what the end result in terms of numbers mean. Being able to see that play so you know what it means when someone talks about clearance efficiency, or contested possessions, or effective disposals, is very helpful at least to me.

772

Which is what we’re talking about in the ham thread. In that clip the chain should have been Clarke, parish, McGrath and then Begley who was the spare outside the contest with space to run and draw another man.

I think the training watchers have talked a bit about moving the ball sideways more before we go forward and this is the perfect example to highlight where initial sideways movement would be more effective than an immediate forward kick.

Moving sideways with forward momentum is something Richmond do very well. Hopefully Caracalla brings that over.

I’d also like to see Merrett being the guy who becomes the outside spare (if taggers allow it) because his ability to change angles with his kicking can be so damaging. To often last year his possessions were in the heat of the congestion and consequently not as damaging. I don’t think he is great at breaking open congested areas fwiw. I think Shiel with his pace and power, parish and McGrath with their hands and stringer with brute force are better in close than Merrett

2 Likes
773

I think it’s a spacing issue with our contested work. We - for ten years - have been way to close to each other the contest.

The example clip shows how much better it would have been if Parish was 2-3 metres wider. Clear space to run decide and execute effectively. By being less than a metre from Clarke as soon as the handball happens the tackler moves to Parish to apply pressure.

With guys like McGrath and Shiel it’s less of a concern as the burn people off for 5m to creat space and they can stay a step or two closer. But for others it should be at the coal face or 2-3 metres back ready to collapse in for the tackle should we loose the scrap for the ball.

We are always too stationary as well.

3 Likes
774

On structure this how I saw it a few weeks ago. There was a significant structure laterally(Rugby) just as much as there was in front and behind the play. It was very much man on man so I didn’t see the benefits but if an opposition team structures differently one of them lateral movers is hopefully loose or has a gap to run forward to either create time to correctly dispose, further run from handball and/or blocking and getting it over the zone. You don’t get them options often or on purpose if the ball carrier stays within the contest.

1 Like
775

https://www.foxsports.com.au/afl/afl-2020-preview-home-ground-advantage-in-afl-analysis-stats-teams-with-best-home-advantage-noise-of-affirmation-what-is-hga/news-story/cebf8140cbca412b746170aeeff07411

THE NUMBERS (Home wins in a season, 2010-19)

9 or more wins = 100% of teams played finals

8 = 91.3%

7 = 50%

6 = 28%

5 = 15%

4 or fewer = 0%

HOME GROUND RECORDS (2017-19)

Richmond (MCG) - 40-5 (89% win rate)

Geelong Cats (GMHBA Stadium) - 22-3 (88%)

West Coast Eagles (Optus Stadium) - 21-6 (78%)

GWS Giants (Giants Stadium) - 19-1-6 (73%)

Adelaide Crows (Adelaide Oval) - 25-13 (66%)

Essendon (Marvel Stadium) - 16-9 (64%)

Port Adelaide (Adelaide Oval) - 22-15 (59%)

Western Bulldogs (Marvel Stadium) - 21-15 (58%)

Fremantle (Optus Stadium) - 14-11 (56%)

Sydney Swans (SCG) - 18-17 (51%)

Collingwood (MCG) - 24-1-22 (51%)

Hawthorn (MCG) - 15-16 (48%)

St Kilda (Marvel Stadium) - 18-1-19 (47%)

Melbourne (MCG) - 16-21 (43%)

Brisbane Lions (Gabba) - 15-20 (43%)

North Melbourne (Marvel Stadium) - 15-20 (43%)

Carlton (Marvel Stadium) - 5-17 (23%)

Gold Coast Suns (Metricon Stadium) - 6-21 (22%)

The other strong home performer is Essendon , who have won almost two thirds of their matches played at Marvel Stadium in the last three seasons; overall, they’ve gone 36-30, so they’re quite a bit better under the roof.

The Bombers will play nine games at Marvel this season - seven home, two away - so that could give them a good base to build off of.

1 Like
776

An analysis that considered who was played against would be a tad more defendable.

1 Like
777

Exactly. Normalise to ELO at the time of playing or something simple.

778

Question for the stat nerds:

In a 22 round season, a 12-10 win loss record would typically scrape you into the top 8.

In a 17 round season, one would think that a 9-8 season may be just enough. From a statistical probability point of view, would this be likely? Or would 10-7 be more reasonable?

If it’s 9-8, Essendon has already achieved greater than 10% of its required wins. Beat Sydney next week, they need 7 wins from a remaining 15 games.

Whilst you don’t want to scrape into finals each year, it appears that in a shortened season like 2020, there is less leeway to lose games. Hence Fremantle would be cursing and Essendon have a very valuable win.

This assumes they will get to 17 games (which I doubt).

779

I think the word is that it may be more than 8 teams to make finals.

780

Then this season will feel even more like an Ansett Cup.

1 Like
781

This season is all about recouping as much of their investment as possible.

1 Like