To summarise, we've three Funds operating in 2012: the Head to Head Fund, the Line Fund, and the Margin Fund, the features of which are summarised below.

Since this will be the first time that MAFL's ventured into a margin market, I thought it might be illuminating to do a little more analysis to understand the historical wagering behaviour of the Fund algorithm I've selected, to bolster the scant details I provided in the previous blog.

Firstly, for the entire period 2000 to 2011, I can report the following facts.

Overall Accuracy of Predictions 

• Margin Fund when it wagers (ie when it predicts a Home team win or draw): 13.7%. (So, unless our winning wagers carry an average price of at least $7.30 - or we can tip at a better rate than historically - we shouldn't expect to make money this year.)
• TAB Sportsbet Predictions: 9.9% overall; 9.5% when the Margin Fund wagers.

Accuracy for Different Combinations of Margin Fund Prediction and TAB Margin Prediction (minimum 30 cases)

• Highest rate: 19% for a Margin Fund prediction of Home team by 10-19 points and a TAB prediction of the same margin range. Such games represent about 10% of the Fund's wagers. (It's hard to imagine that we would've lost money on those games had we been wagering.)
• Other success rates of 15% or higher: 17% for Fund prediction of Home win by 50-59 and TAB prediction of Home win by 40-49 points (7% of wagers); 16% for Fund prediction of Home win by 40-49 and TAB prediction of Home win by 30-39 points (another 7% of wagers).
• Lowest success rate: 9% when the Fund and the TAB both tip 20-29 point Home team wins, or when the Fund tips a 40-49 point win by the Home team and the TAB tips a 20-29 point win. Combined, such games represent about 13% of wagers.

The Margin Fund will not be wagering when the CNN_2 algorithm tips an Away team win since its accuracy in these cases drops dramatically to just 7.2% across the period 2000 to 2011, only a bit more than one half of its accuracy when it predicts Home team wins.

This decline in performance is not due solely to the relative ineptitude of the CNN_2 algorithm when predicting Away team victories. The TAB bookmaker also suffers a reduction in accuracy for games where it tips an Away team win, though its accuracy falls only from 10.9% to 9.3%.

For the 2011 season considered alone we find the following:

Overall Accuracy

• Margin Fund: 14.3% (up 0.6% points)
• TAB predictions:  overall 13.0% (up 2.1% points); for games in which the Margin Fund wagers 13.3% (up 3.8% points)

The good news is that the Margin Fund has a higher success rate than the TAB in those games where it wagers whether we look at the entirety of the 2000 to 2011 seasons (13.7% beats 9.5%) or just the 2011 season alone (14.3% beats 13.3%). 

Finally, note that the strategy of wagering only when CNN_2 and the TAB agree on the predicted victory margin, whether that margin be for a Home or for an Away team victory, would have been highly successful. Across the 2000 to 2011 seasons such a strategy would have predicted the correct margin 11.8% of the time and, for the 2011 season considered alone, it would have predicted the correct margin an astonishing 18.5% of the time. If wagering was constrained only to those games where both CNN_2 and the TAB predict a Home team victory, the success rate lifts to 13.3% for the entire 2000 to 2011 period and to 19.2% for the 2011 season alone.

One of the statistics we'll track this season is the accuracy of CNN_2 margin range predictions when they coincide with the TAB's versus when they do not.