With the Line Fund snaring four from six wagers to jump by 4c, and the Head-to-Head Fund winnowing a small profit from a pair of successful wagers having tantalised with two more, it felt as though we deserved more this week for our frenzied efforts. But, the Margin Fund's none from eight knocked 10c off its price and was enough to hand back to the TAB Bookmaker all but 0.3c of the profits contributed by the two other Funds, leaving the Recommended Portfolio virtually unchanged and now down by an even 2c.
Next week the Line Fund doubles its bet size while the Margin Fund drops its bet size by 20%. On the strength of the two Funds' recent performances its hard to do anything but applaud those adjustments.
TIPS AND PREDICTIONS
Only two favourites - Carlton and West Coast - lost this weekend, which made for a relatively high 6.0 average score for the Head-to-Head Tipsters. The nine Tipsters currently occupying the nine highest places on the Leaderboard all picked seven games correctly, which proved to be the best that any of the MatterOfStats Tipsters could muster. Low score was four from nine and belonged to Shadow, which shuffled it further down the ladder leaving it now 14 tips off the pace. Home Sweet Home still props up the table, some 17 tips behind the leader, Bookie_9, which remains a lone tip ahead of Combo_7 and a further half tip ahead of Bookie_3.
Win_7, Win_3 and Bookie_9, who each registered seven correct predictions this week, also registered the week's three best Mean Absolute Prediction Errors (MAPEs). Win_7 recorded the best MAPE, its 37.2 points per game effort seeing it finish the round atop the Leaderboard, now 23 points ahead of Win_3.
The all-Predictor average was 40.7 points per game, dragged down somewhat by C_Marg's 47.0 MAPE result. C_Marg remains, however, one of only four Predictors with a season-long profitable line betting record, the other three being Win_3 and Win_7 from the head of the table, and H2H_Unadj_3 from nearer the foot.
C_Marg's sibling, C_Prob, continues to impress amongst the more-seasoned Probability Predictors, lying third, just ahead of Bookie-LPSO. Things could have been very different for the Head-to-Head Predictors, C_Prob in particular, this week had the Giants got home over the Hawks, an outcome which seemed very plausible for much of that game. C_Prob stood to lose about 6.7 bits, equal to about two-thirds of its entire probability score for the season to date, while Bookie-OE, Bookie-RE and Bookie-LPSO each stood to lose between about 3.1 and 4.6 bits. The Log Probability Score truly does punish probability assessments near certainty for teams that subsequently lose.
The round's best Probability Score was recorded by Bookie-LPSO, but it, Bookie-RE, C_Prob and WinPred all registered very similar, positive scores.
The Line Fund algorithm, however, registered a negative probability score this week, its second in a row but only its fourth for the season.
Despite having three more chances to choose correctly compared to the past three rounds, none of the Margin Predictors selected the correct bucket in more than a single game this weekend.
Only six Predictors managed even that, and for none of them did it turn a previously loss-making season into a profitable one.
SuperMargin prediction really has been difficult this year, as evidenced by the fact that none of the Margin Predictors has guessed correctly more than 11% of the time, 12% if we consider their predictions only when they foresee a Home team win or draw. It's impossible to make money with percentages this low when you're, typically, wagering at prices around $7.