2014 - Round 14 Results : Make That Five

By Sunday afternoon, Investors had made enough money during the round that neither Saints nor Demons could truly bother them. 

In the end, those Saints and Demons did take away the profit that the Giants had delivered earlier in the day, but that still left us with a 9c profit from the weekend. Contributing to this result was the Head-to-Head Fund, which gained just under 5c on a 2 and 1 record, and the Line Fund, which leapt 17.5c on a 5 and 1 record that took its cumulative tally since Round 8 to 21 and 7.

The Margin Fund was, again, the only Fund to lose money, but it did go within a point of landing the Hawthorn result - slipping narrowly into the dreaded Chasm of Despair - so once more I'll give it a pass. It'd seem churlish not to.

Across the season now we have the Recommended Portfolio up by just under 16c having strung together five consecutive weeks of profitability, with the Head-to-Head Fund now up by about 11c, the Line Fund up by 44c, and the Margin Fund down by 47c.

Long may the Line Fund's reign continue (and may regression to the mean stay its course). We still have four more rounds where the Line Fund's bets are at their maximum level for the season.


Bookie_9 extended its lead atop the Head-to-Head Tipster portion of the Leaderboard this week, its 7 from 9 performance taking it to 85 from 117 (73%) for the season, two tips ahead of Combo_7, and three ahead of Bookie_3 and the three WinPred-based Tipsters. 

All up, the Head-to-Head Tipsters averaged 6.3 from 9, with four of the Heuristic Tipsters - Silhouette, Shadow, STM II and EI II - producing the round's best result of 8 from 9. Worst were the 5 from 9 results for EI I, STM I and FTS.

The round's best margin prediction performance belonged to Win_3, its 18.1 Mean Absolute Prediction Error (MAPE) consolidating its second-placing on the Leaderboard behind Win_7, which it trails now by less than 4 points. 

MAPEs were generally good this week, the average coming in at 20.2 points per game and the highest MAPE belonging to Combo_NN_1 at a nonetheless respectable 23.3 points per game.

Twelve Predictors now have profitable records on Line betting, while only one has tipped at worse than chance. Win_7 and Win_3 have particularly impressive Line betting records, having selected the correct team about 60% of the time. Further evidence of their outstanding predictive performances so far this season comes from the fact that Win_7's margin predictions have been with a single goal of the actual outcome 18% of the time, and Win_3's have been the same 16% of the time.

Even records as good as these have not been sufficient to generate profitability in the SuperMargin market for these or for any other Margin Predictor, had wagers been made on all their predictions.

Only by ignoring games where the prediction implied an away team victory could a profitable strategy be extracted from any of the Predictor's suggestions. Such a strategy, if applied to the musings of ProPred_3, ProPred_7, H2H_Adj_7 or Combo_NN_1, would have generated modest positive returns.

Returning to the Leaderboard to look at Head-to-Head Probability prediction reveals an unchanged Top 3, with Bookie-OE still leading, Bookie-RE still in second, and WinPred retaining third. C_Prob, after returning the round's best probability score, has grabbed fourth place, however, relegating Bookie-LPSO, last year's end-of-season leading Predictor, to fourth.

The Line Fund algorithm had its second-worst probability score performance all season, better only than its Round 11 score, dragged down most heavily by its assessment of the Hawks' chances on line betting as being only 18%. The Log Probability Score metric really does punish misplaced overconfidence