St Kilda's two final scoring shots weren't good for my mental state on Sunday afternoon, coming as they did with the Swans holding a 61-point lead, barely sufficient to cover the 57.5 points start we were contractually obliged to offer the Saints as our part of the line bet. Had either scoring shot been a goal rather than, as both turned out, behinds, Investors would have finished the round with a much more substantial loss (and my Sunday dinner would have tasted much worse).

As it was - a fact to which I was oblivious at the time - those two behinds secured not only our line bet but also morphed one of our SuperMargin wagers into a collect as well, a combined result that could only be achieved if Sydney won by exactly 58 or 59 points.

In all, we wound up losing just 0.8c on the round, leaving us up by 8.4c on the season. Across the weekend, the Head-to-Head Fund lost both of its wagers, dropping 3c in doing so to finish the round up by about 12c, the Line Fund went 2 and 2 to shed half a cent to 4c, and the Margin Fund landed two wagers from 10 to jump 5c and face the next round up 24c on the season.

That's three losses in four weeks now for the Recommended Portfolio and solid vindication of the decision to ratchet our wagering back in the latter part of the season. Results really do feel like something of a lottery at the moment.

TIPSTERS AND PREDICTORS

Seven favourites won this week and most Head-to-Head Predictors duly recorded seven correct predictions, with none managing a higher score but eight scoring lower. Worst was Follow The Streak with a score of just three. There was, then, no change then at the top of the MAFL Leaderboard.

(Regular readers might notice some changes to the format and content of the Leaderboard this week and to the Fund Summary in the navigation bar. These were necessitated by the malfunctioning of my sparklines add-in for Excel. At this stage I think it's terminal ...)

Most of the changes I've made to the Leaderboard are cosmetic in nature - adding the "Gap" column in the section on the left to replace the stacked barcharts that were there previously, deleting the columns of percentages and reconstructing the stacked barchart in the middle of the Leaderboard to now incorporate those percentages - but the ROI column is new. It charts the ROI that would have accrued to anyone wagering on a Margin Predictor's line betting predictions and securing a price of $1.90 for each wager. As you can see, six Predictors have a positive ROI on this basis, though only the two RSMP's ROIs are significantly non-zero.

On the basis of the mean absolute prediction error (MAPE) metric, Bookie_LPSO slipped back into 3rd place this week after recording a 25.7 MAPE, relegating Combo_7 to 4th place with its 26.4 points per game performance.

The Round's best MAPE (24.9) went to RSMP_Weighted, which allowed it to marginally extend its lead over RSMP_Simple at the top of the Leaderboard. How amazing it is that we're now 21 rounds into the season and we have five Predictors with sub-27 MAPEs, and all Predictors with sub-30 MAPEs.

Bookie_RE recorded the best probability score of the round amongst the Head-to-Head Probability Predictors, a gnat's gnat better than Bookie_LPSO and not that much more superior to Bookie_OE and WinPred. In fact, so similar were the probability scores of all Predictors this week that their ordering on the Leaderboard remained unchanged.

The Line Fund algorithm produced a slightly negative probability score for the round this week, its second such score in a row and its third in the last four rounds.

SUPERMARGIN PERFORMANCES

You'd have been best-served this week using the TAB Bookmaker's line betting handicaps as the basis on which to enter the SuperMargin market since this strategy would have landed you two successful wagers, one better than you'd have done with any other of the MAFL Margin Predictors.

ProPred_3 remains the Predictor with the best season-long performance, having now selected the correct bucket on 26 occasions, which is once more than Win_7 and twice more than ProPred_7 and Combo_NN2. Those four Predictors, along with H2H_U10 and H2H_A3 are the only Predictors showing a positive ROI for the season.