With the home-and-away season behind us, it's time for review and reflection on a variety of fronts.
Wagering Returns By Team
Firstly, let's look at which teams are to blame for our wins and losses in wagering. In the head-to-head market we've wagered on every team except three so far this season. Those three, understandably, include the Dogs and the Suns, but also include the Hawks, who finished the home-and-away season 17 and 5 and so must surely have offered value at least once this season.
Six teams produced a net profit for us, best amongst them the Tigers, Cats and Roos who, collectively, lifted the Head-to-Head Fund by about 25c. Nine teams cost us money, most notably the Saints, who we wagered on five times for three wins and two losses, but for a net 9c loss, and the Swans, who we wagered on only once but for an 8.5c loss.
GWS have been the major Line Fund contributors, winning on six of the eight occasions on which we've favoured them with our fiscal attention. Eight other teams have been net profitable for us, including Sydney (+9c), West Coast (+8.5c), Gold Coast and Hawthorn (+8c each), and Richmond (+7.5c). We've lost money on the ten other teams, including double-digit losses on the Blues and the Pies (16c each), Port Adelaide (15c), and the Saints (11c). Essendon is the team on which we've wagered most often, the ten bets on them resulting in five wins and five losses for a net 2.5c loss.
The Margin Fund has made a net profit from wagering on only five teams, very much foremost amongst them Adelaide from which it has netted almost 32c having landed three of nine wagers - a startlingly high proportion of SuperMargin wagers. Collingwood, Geelong and St Kilda have been the major value destroyers for this Fund with, amongst them, 25 unsuccessful wagers from 25 attempts.
Flipping the view around and looking at our wagering on the basis of the team we were wagering against yields the following:
In head-to-head wagering only the Cats can lay claim to being profit producers both when wagered on and when wagered against. Both Fremantle and Melbourne, however, have been responsible for losses both when wagered on and when wagered against. Sydney has been the team we've most often wagered against in head-to-head betting this year. We've done this five times for three wins and a net gain of almost 4c.
The Line Fund has many more dual-profit and dual-loss makers. The Suns, Giants, Tigers, Eagles and Dogs have all been profitable when wagered on and when wagered against, while the Crows, Blues, Pies, Dons, Dees, Port, and Saints have all been loss-makers when the games in which they've been involved have been the subject of a line wager. Hawthorn and the Roos have been the teams we've most often wagered against on line betting this season, in Hawthorn's case to disastrous effect, costing us over 25c.
Only the Crows have produced SuperMargin wagering profits both in games where they've been wagered on and wagered against, but the Lions, Blues, Pies, Dees, Eagles and Dogs have all been responsible for net losses when they've been the teams wagered on or wagered against. GWS is the team we've most often wagered against in the SuperMargin market, but too often we've punted that they'll lose by far less than they've been expected to, and too often they've exceeded expectations in the very worst way.
Wagering on the Heuristic Tipsters
We've often found in the past that level-staking the Heuristic Tipsters has been profitable, especially if we only start wagering from Round 6 onwards. This year, some of the Heuristic Tipsters have again produced profits based on following this strategy, but only if you'd done so commencing in Round 1, not Round 6.
Kelly-Staking versus Level-Staking
The Head-to-Head Fund has Kelly-staked throughout this season while the Line Fund has Level-staked. These wagering policies have produced ROIs of -7.2% for the Head-to-Head Fund and 0% for the Line Fund. You might be curious to know what ROIs would have been produced by Level-staking the Head-to-Head Fund (on the basis of the unadjusted H2H probabilities) and by Kelly-staking the Line Fund (on the basis of the Line Fund algorithm probabilities).
Well, Level-staking the Head-to-Head Fund would have resulted in a +15.2% ROI (even higher if we'd also removed the prohibition on wagering on teams priced under $1.50), while Kelly-staking the Line Fund would have resulted in a -0.7% ROI.
Tipping and Predicting: 2011 versus 2012
Lastly, let's review the comparative performances of the various MAFL Tipsters and Predictors in 2011 versus 2012.
We can see firstly in head-to-head tipping how well almost every Tipster has done in both years, with 16 Tipsters recording accuracy rates of 70% or higher back-to-back. Summed across the two years, BKB has a one tip lead over Combo_7, with Bookie_3 another half tip back, and H2H_Unadj_10 and H2H_Adj_3 yet one more half tip behind Bookie_3.
The general demise of Home teams is also in evidence, with Home Sweet Home finishing last in both seasons with an accuracy of only around 56%.
On margin prediction, Bookie_3 and Combo_7 have filled the top two places in both years, with various other Predictors filling the next few spots. Only H2H_Unadj_10 has finished in the top six in both seasons. Bookie_3 is the only Predictor to finish with a sub-30 MAPE in both seasons.
Seven of the Margin Predictors have recorded better than 50% line betting records in both seasons, best amongst them ProPred_3, which has a two-season combined 56% record, comfortably better than chance.
Probability Scoring has also be atypically high in both of the past two seasons, with averages of around +0.150 being more typical of earlier seasons. This year, proportionately, the non-TAB Probability Predictors have finished a lot closer to the TAB than they did in 2011. WinPred, for example, has an average probability score that is over 90% of the TAB's average; last year the nearest Predictor at this point in the season was H2H_Unadj and its score was only about 84% of the TAB's.
The Line Fund algorithm has also had a much better year in 2012, its average probability score per game coming in at -0.029, well up from last year's -0.048.