I should probably have noticed this sooner, but last year was quite a profitable year for blindly wagering on Home Teams. A gambler who level-staked the AFL Designated Home Team in every game in the head-to-head and in the line market would have recorded an 8.4% ROI on his or her head-to-head wagers and a 4.1% ROI on his or her line wagers.
On the one hand this makes me more comfortable heading into the upcoming season armed only with monogamously Home team models, but it also makes me less impressed by the fact that these models show a profit for the 2010 season. More on this in a moment.
First, though, a little history on the general efficacy of backing home teams head-to-head. Across the seasons 2006 to 2009 - which are the four other seasons during which MAFL was active and for which, as a consequence, I have bookie data that I trust - level-staking AFL Designated home teams head-to-head would have been profitable in only one of them, 2007, when such a strategy would have yielded a modest 2.5% ROI. In the other three seasons, negative ROIs of around 3% would have been recorded. Set against that background, the 2010 profit seem all the more unusual.
The slightly more refined approach of wagering on home teams only if they're underdogs, which for the current exercise I'll define as home teams priced at greater than $2.20, would have been more generally lucrative across the five most recent seasons, especially in 2010 where the ROI would have been almost 11%, but also in three more of the remaining four seasons in which ROIs would have been in the 1.5% to 4% range.
So, some five years after I first noticed it, level-staking home team underdogs still appears to be a generally sound strategy. The original observation was based, I think, on the data for seasons 2004 and 2005, which show ROIs for home team level-staking of around 5%, and ROIs for home team underdog level-staking of around 7% - which would surely have been enough to attract my attention then as now.
Returning then to an assessment of the models I've chosen to use in season 2011, it seems appropriate to adjust their 2010 performance for the fact that I've steered them from wagering on away teams. A little simulation shows that any strategy that randomly wagered on home teams head-to-head in 2010 would, on average, have produced an ROI around 8.4%. In fact, if you take the actual wager amounts that our head-to-head model would have recommended last year and randomly assign these amounts to games so that, for example, the 3.5% wager on Geelong v Hawthorn in Round 15 is instead applied to the Richmond v St Kilda clash in Round 11, on average the result of such random assignment is about an 8.4% ROI.
The good news is that both the head-to-head and the line model comfortably outperformed a comparable naive, level-stake all Home teams model. Our head-to-head model recorded an ROI of 18%, which is more than double the ROI for the naive head-to-head model, and our line model recorded an ROI of over 14%, which is more than three times the ROI for the naive line model.
One perhaps slightly unusual measure of how unlikely is that 18% ROI is that in less than 1-in-4 of the simulations where I randomly assigned the head-to-head model's wagers to 2010 contests was the resulting ROI greater than 18%.
Better still, both models recorded positive ROIs for 2009 when their equivalent naive counterparts recorded ROIs of negative 3.3% (head-to-head) and negative (1.7%). So, even when the general environment was mildly hostile to random home team wagering, our models managed to eke out a return.
Still, none of this matters unless we see a positive return in 2011 (and even then, of course, we can never be certain that I haven't just been lucky).