For Investors, there's not a lot of news to add to last night's other than to lament that we had so little riding on the wings of the victorious, underdog Saints. As I shared with one Investor near game's end: form is temporary, but MARS Ratings are semi-permanent.
The 1.3c added by the Head-to-Head Fund on the back of the Saints' win kicked up Overall Portfolios by only 30% of that or about 0.4c, but this was enough to drive all Portfolios fractionally into the black for the season. They now stand at $1.003c.
With the Saints making it four underdogs grabbing the competition points this week, Head-to-Head Tipsters collectively turned in a season-low average of just 4.8 from 9 for the round. Some of the season's stragglers - HSH, EI I and EI II - recorded the round's best performances with 6 from 9, while CNN2 was responsible for the worst performance with just 3 from 9. Thirteen Tipsters remain joint leaders on the MAFL Leaderboard, all on 50 from 63 (79%).
Margin prediction also suffered this week, due partly to the prevalence of upset victories, but due also to the fact that the victory margin in five games exceeded 6 goals. The all-Predictor average of 36.95 points per game was the second-highest of the season. Bookie_3 recorded the round's best MAPE of 33.57 points per game, while Win_3 recorded the worst at 38.29 points per game. Bookie_9 continues to lead all-comers based on season-long results, ahead of Combo_7 and Bookie_3.
Weaker margin prediction performance came with poorer line betting results too, but the 5.2 from 9 all-Predictor average is still somewhat better than chance, and it remains the case that every Predictor bar CNN_2 has a season-long record exceeding 50%.
What's more, in four of the week's matches at least one of the Margin Predictors selected the correct SuperMargin bucket. In the Melbourne v Hawthorn game it was CNN_2 (unfortunately based on predicting an Away team victory) and Bookie_3 that had it right; in the Richmond v Sydney game it was ProPred_3 and ProPred_7; in the Brisbane Lions v Collingwood game it was ProPred_3 again; and in the Fremantle v Port Adelaide game it was Bookie_3 again.
Even with the successes of these Predictors this week, amongst them only CNN_2's predictions have been notionally profitable across the entire season so far (see table below). The season-long performances of Win_3 and Win_7 also remain worthy of comment: respectively they've selected within a bucket of the correct result in 27 and 26 games, but they've been exactly right in only 4.
For the first time this season, all four Head-to-Head Probability Predictors recorded negative scores for a round, though in the case of the TAB Sportsbet bookmaker it was only narrowly sub-zero. The two Head-to-Head Probability Predictors now lead, comfortably ahead of ProPred in 3rd.
The Line Fund algorithm had another round in which its probability score across the nine games was negative, its fourth such round of the season so far. Previous analyses have shown that Investors can make money when the Line Fund is spouting negative probability scores, but not when it does so excessively; some even of improved calibration would be calming.