Paul Roos on the 'modern game'

[cruiser]

This is sure to spark some debate here: http://realfooty.com.au/news/news/co...86.html?page=2

In particular this quote, with a reference to Demetriou's 2005 ugly Sydney comments:

"The good thing about what has probably transpired since those famous comments a couple of years ago, I think people have tried to understand the game a bit more but there is still some of that, 'kick it long, get it in quick, contested mark' [thinking].

Surely the best strategy to beat a flood is to get the ball down quickly to your forward line before your opposition has time to get its players back to flood? Roos used it himself in the second half against the Evils in round 1 and it almost carried Sydney over the line to a win.

[NMWBloods]

In The Age yesterday I think there were the results of an analysis of nearly 800 games of football. This was done on a similar basis to Michael Lewis' famous Moneyball analysis of baseball. It highlighted key measures that were important and those that had little influence. As I recall long kicks were highly correlated with success and tackling wasn't. Can't recall the others.

Ah - here it is.

What matters most:

Accurate long-kicking, bounces, centre clearances, knock-ons.

What is not important:

Umpiring decisions, ruck contests, tackling.

[giant]

Stewart and his team analysed 738 of the 740 AFL games played between 2002 and 2006.

The two games they missed must have been the 2005 & 2006 GFs.

[Legs Akimbo]

I was really interested in that local version of the moneyball analysis. Not sure how they did it because they don't explain what the coefficients mean in the article I read. But I expect if it is a multiple regression analysis i.e. where the coefficient is the weight for each variable in the equation in terms of the proportion of variance explained in the dependent variable 'winning'. Of course they don't show the error term in the equation in the newspaper article and therefore there is no way to assess whether it is a good model or not.

The explanation of the model was certainly simplistic. i.e. tackling accounts for 'x' weight on winning. A more sophisticated analysis (e.g. structural equation modelling as one example) can take into account cross effects. e.g. tackling on its own might not a strong relationship with winning, but in combination with effective hitouts and effective short kicks, it is much more effective (i.e. almost a 'niche' style of play). In fact, I think this is the reality of the Sydney swans game plan.

I think it was last year I did a simple factor and cluster analysis of team statistics. The result indicated that there are different styles of play. The two styles of play with the strongest winning percentage were (i) long kicking open style of play (e.g. the north melbourne of old) and (ii) the tight contested style of play (e.g. the Sydney method - winning contested footy, tackling, 1 percenters etc.). I think this understanding more strongly reflects the reality than simply citing that these are the statistics associated with winning.

Then again, these guys are complete propeller heads so I presume they look at that. I probably should read the actual source document to get a better idea of their model.

[liz]

Agree with you Legs that multiple regression approaches can sometimes turn out models with limited meaning. I guess you can put interactive terms in there but with a hit and miss approach of trying almost every combination, chances of finding the true interactive terms are low, unless you start with a theoretical view and then try to use the data to prove you are right (or wrong).

The idea that tackles aren't important is counter-intuitive and may be partly explained by the fact that real data of AFL games rarely shows much difference in tackles. That is because every team knows they have to tackle, and when teams less inclined to do so come up against teams - like Sydney - which do, they invariably up the ante in the tackling department. If real data doesn't display much variation in a specific variable, it is unlikely that a MR model will find that that variable has much of an impact.

If any game data existed where there was significant differences in tackle counts, I suspect it would be found to have important effects.

One other issue with tackling is that you can't always take the raw numbers at face value. If a team is continuously second to the ball and chasing their men, chances are they will have a moderately high tackle count at worst. However, if a team is winning clearances and hard ball gets and STILL has a high tackle count, you can reasonably infer that their overall pressure skills have been high.

[reigning premier]

Look, this is all well and good..... But aren't we getting just a little bit too scientific here????.....

As we always say at the game...

Kick...then Catch... Kick... Then Catch.... repeat.

Pretty straightforward stuff really....

[stellation]

Davis > Kirk

[liz]

Look, this is all well and good..... But aren't we getting just a little bit too scientific here????.....

As we always say at the game...

Kick...then Catch... Kick... Then Catch.... repeat.

Pretty straightforward stuff really....

Except that is mostly what St Kilda and Hawthorn spent two hours doing last weekend...

[reigning premier]

Davis > Kirk

well der.....

[reigning premier]

Except that is mostly what St Kilda and Hawthorn spent two hours doing last weekend...

Everyone seems a bit down with the whole "maintaining posession at all costs" strategy. I for one think it's a legitimate part of the game. We certainly do plenty of it. And why the hell not??? The opposition can't score if they don't have it. And when a game is played in this way, it's always the more patient and disciplined team that comes out on top.... i.e Us!!

[hammo]

Football would be a much easier game if there was no pressure and no risk of being tackled. Skills under pressure - and how teams apply that pressure - is the cornerstone of the modern game. Every AFL footballer could hit a target with no pressure, the difference is when, as Sydney do, immense pressure is applied around stoppages and general play.

Only the very best teams can execute their skills when placed under enormous pressure.

The study is rubbish.

[NMWBloods]

One other issue with tackling is that you can't always take the raw numbers at face value. If a team is continuously second to the ball and chasing their men, chances are they will have a moderately high tackle count at worst. However, if a team is winning clearances and hard ball gets and STILL has a high tackle count, you can reasonably infer that their overall pressure skills have been high.

I think that is the key thing on tackling. The numbers are not significantly different and if you have the ball your tackling will tend to be a little lower again.

Where tackling may be important is if the number is very low in which case you may have a problem.

So I see tackling (sort of like ruckwork) as a minimum thing to keep you in the game, but the marginal benefit of additional tackles is small and will not win a game.

In 2007 Sydney, Geelong, Adelaide and West Coast have negative differentials on tackles relative to their opponents.

[Legs Akimbo]

So I see tackling (sort of like ruckwork) as a minimum thing to keep you in the game, but the marginal benefit of additional tackles is small and will not win a game.

Hygiene factor