Model version 9 over the years

As mentioned in earlier blogs the Carlo Monty Model version 8 did not perform as well as hoped in the previous two seasons and has been re-examined with the claret ity (or more accurately Pinot Noir) of a couple of weeks of tweaking.  Where the previous models have been based on the data from 2000 where I felt that the grounds were all of similar quality and the training of players could be roughly considered the same I finally bit the bullet and uploaded the entirety of AFL/ VFL data and crunched the numbers.  I have made a few home ground advantage differences for different eras - I am old enough to remember the sprinklers being left on at Moorrabin "by accident" and Kennedy's commandos running around the cigar shaped ground that was Glenferrie Oval.

I must admit that wit hall of the tweaking the shot model algortihms that I have relied on for the past decade stubbornly bounces around the 60-70% accuracy



The one breakout above 80% accuracy being in 1900 - not even my grand father was alive to give the SP bookies a run for their money back then.  Model 9 results shown above are the result of my current optimising over the entire history (and my previous good 2011 result is the one season in the 2000's that drops dramatically and is the current focus of what else can be tweaked).

Looking at the 2000's I ranked each team for model accuracy for the games in which they appeared:



The Tigers have not only given their fans much grief with numerous 9ths and the Roos are not much better.  The two newest teams on the paddock have been easier to predict an outcome in the games they appear in (by assuming they will loose mainly!)

With inspiration from MAFL's analysis of home ground advantage I am trying to see if I can modify the model legitimately to squeeze out a few more accurate predictions with games at the MCG and Etihad/ Docklans for teams that although they may be their home ground but their opposition may have played as many or more games there in the recent past.

Hopefully 2014 will be more like 1900! We live in hope.