Shots for and against

While watching my tip of a North win turn to another donut I thought that I would have a look at some of the graphs in my model.  I decided to have a closer look at the shots for and against (5 point moving average) for each team compared to the distribution of data from the start of the 2000 season up to the end of round four 2014.



The 5364 scores resulted in an average points per scoring shot (excluding out of bounds and including rushed: i.e. the scoreboard totals of goals and behinds) of 3.67; with a maximum of 5.375 and a minimum of 1.625; standard deviation of 0.531.  All fairly straightforward excel number crunching.



With a little more excel work I converted this data set into a cumulative distribution plot and have highlighted with two vertical lines the 33 and 66% (3.432 and 3.895 respectively).









Now taking the broad brush approach I consider a team has a good strong attack if they score more than 4.0 points per scoring shot and similarly if their opponents score more than 4.0 points per scoring shot against them then they have a leaky defence.  A tight defence concedes no more than 3.40 points per scoring shot and an attack under a lot of pressure when trying to score would also be less than 3.40.



Now of course these number bounce around each week for each team depending on who they play, the weather etc.  However, continuing with the broad brush analysis it is fairly easy to also use excel to calculate the moving average for each team (although in my case it does need a lot of look up tables and the use of the very handy function vlookup an awful lot of times).  The five point moving average for each team's offence and defence going into round five is shown in the table below:







I have highlighted the cells which are in the top or bottom third compared to the 2000-2014 data set.  Now the table shows some things that we/ the general public perception have in mind: that Fremantle have a very strong defence and that Sydney also play a defence orientated game.  Also that teams generally have scored easily against the Giants and that Melbourne have not been scoring very much.



However, I also think the table highlights a few things that are not commonly discussed: notably that the Bulldogs have a far more free flowing game with a leaky defence and an above average attack.  That Richmond and North are both "middling" teams as they are neither noted for their attack or defence.  Collingwood an Melbourne have a similar "pattern".



The table also highlights that to "be green" in both shots for and shots against is rare - only Hawthorn achieve this at the moment with Geelong only a couple of straight kicks away.  Which justifies the game of the round status on Monday where I hope that the quality of the game means that I will stay focused on the game rather than on the excel number crunching or the twitter commentary.

Blowout Round

I think the table says everything that needs to be said regarding how expert the expert tipsters are. (blue dots show the published margins from the papers and a few internet tipsters including my own; red dots actual results for round 4)

Experts and the Crowd - Margin predictions

Having taken the plunge and signed up for the Footy Mathis Institute Margin Tipping competition (first prize: bragging rights; last place potential ridicule) I have decided to try and understand margins a bit more.  Previously I focused on first pass the post win/ loss tipping but have used the margin predictions from my model to estimate the odds which in turn I use to compare where I have a difference of opinion to the bookies.



So aside from looking at my fellow Footy Maths tipsters and the collection of like minded internet bloggers (Footy Forecaster, Matter of Stats and Tip Bet Pro) I have started to pay more attention to the "Expert" tips in The Age and the Herald Sun (ignoring the Village Idiot etc).  Having entered the numbers for the first three rounds I am surprised at how poor the margin tipsters are - and how very poor my margin tips have been.  Of the 1761 margin predictions a total of 21 have been spot on and 231 within 6 points (Notably Footy Forecaster picked two correct margins in Round 2).  When I plot the cumulative percentage of the difference between the predicted and actual margins I was very surprised to see an essentially straight line (with 50% within a 5 goal margin and the other 50% up to about 100 points off the mark!)







Moving onto the second item of interest is how many of each rounds actual results fall outside the range of predictions.   Looking at round 1 for an example it is obvious to all that GWS beating Sydney was very unexpected.  But Collingwood being opened up by Fremantle was as unexpected based on the distance from the actual result to the nearest prediction from the experts.  None of the experts came close to predicting Essendons victory over the Roos or the Bulldogs loss to the West Coast.  Add to that the result of Gold Coast over Richmond almost falling outside the experts prediction window and you really need to re think how expert the collection of experts are.







rounds 2 and 3 do not fair much better:



Round 2: Port, Sydney and West Coast all outside the experts range.





Round 3

Hawthorn, Adelaide, Gold Coast and Essendon are outside and Bulldogs only fall in as I assigned Murphy a ten point victory margin for his team (probably should remove this from the analysis)









 

So all up twelve games actual results fell outside of the range of predictions from 75 Experts.



I am intending to look at the relationships with average margin predictions, bookies odds and whether the tipping accuracy improves over the season over the coming weeks but to date I think it is fair to say the expert opinion should be called the expert guess.