It's a Zen thing. Winning by losing; succeeding by failing; being right by being wrong; profiting by making losses. Either that or football betting is subject to a version of the Heisenberg Uncertainty Principle such that you can only understand the statistical characteristics of AFL or accurately estimate teams' true victory probabilities, but not both. In which case, five years of analysing and writing about the former has been irretrievably detrimental to the latter.
That's how it seems tonight, anyway, after a round where we lost most heavily on the lone losing outright favourite and on the outcome that bookies had as a coin toss.
In total, this week Investors won 3 from 5 head-to-head wagers and 1 from 3 line wagers, but they lost the second-largest head-to-head bet (on Carlton) and the largest line bet (on Richmond) and so dropped another 9.7c. Investor Portfolios are now at 75.6c.
It was, as I've hinted, yet another week for the favourites, with six of seven outright favourites taking the competition points. In the remaining game, where Richmond and the Dees were joint-favourites, BKB gets the win because it follows Consult the Ladder when there's no clear favourite and so tipped the victorious Dees.
That means that BKB tipped seven from eight, as did a number of other Head-to-Head Tipsters. Most other tipsters did well too, so much so that the average score for the Round was 6.3 from 8, the 2nd-highest average this season. The last four weeks have, therefore, produced the 5th, 1st, 3rd and 2nd highest all-tipster averages for the season. If only more of the winners had been Home teams - only 18 of 32 of them have been.
Combo_NN_1, which also bagged 7 from 8, still leads the Head-to-Head Tipsters ladder, on 81.5 from 109 (75%), one tip ahead of BKB and Bookie_9 on 80.5 from 109 (74%). Home Sweet Home had yet another poor round, scoring just 3 from 8 to leave it at 61.5 for 109 (56%) for the season.
Collectively, the Margin Predictors recorded the season's 6th-best round with an all-Predictor average MAPE of 28.71 points per game. Best was Combo_NN_2 on 25.26 points per game and worst was ProPred_3 on 31.07 points per game. Bookie_3, with a season-long MAPE of 27.68 point per game, now heads Combo_7 by 0.9 points per game. Five Predictors have sub-30 MAPEs and Combo_NN_2 remains as the Predictor most likely to predict a margin within 2 goals of the actual result.
Using the Margin Predictors for line betting would have produced, on average, 4.6 correct predictions this round, the 3rd-highest average so far this season. As the season progresses, almost all Margin Predictors seem to be being drawn towards a 50% record that denotes no-better-than-chance performance. Bookie_3 remains as the only exception, with a season-long 55% track record, and even it this week managed just 4 from 8.
It was yet another good round for the Head-to-Head Probability Predictors as all five of them registered positive probability scores for the fourth successive round. Best once again was the TAB Sportsbet bookmaker, though this week he was only just ahead of ProPred. The ordering of the Head-to-Head Probability Predictors remains unchanged.
The Line Fund algorithm had a poor round, however, hurt by line betting probability estimates of 24%, 31%, and 26% for Gold Coast, Melbourne and the Swans, respectively.