Matter of Stats

View Original

2014 - Round 4 Results: Well That Won't Pay the Bills

Save for a couple of very comfortable line betting wins on Port Adelaide and Geelong, Round 4 offered Investors absolutely nothing to cheer about.  

Three of four head-to-head bets were near comatose by half-time, while the fourth, on the Giants, at least showed genuine promise until early in the final term. The four other line wagers were also manifestly meritless not all that long after the first centre-bounce, and the 10 SuperMargin bets reached unsaleable status well before the final siren provided the TAB bookmaker with the legal right to move them to the Win column. Practically, he'd had them there much earlier.

Losses, were then recorded by all three Funds, 3.4c by the Head-to-Head Fund, 2.8c by the Line Fund, and 12.5c by the Margin Fund. Combined, that translates to a 4.9c loss for the Recommended Portfolio, plunging it into deficit for the first time this season. It's now down by 0.7c and the only constituent Fund in profit is the Head-to-Head Fund, which is now up by 5.8c.

Let's just draw a dignified curtain around that week ...

TIPS AND PREDICTIONS

You'll recall that the levels of disagreement amongst the Tipsters and Predictors was relatively high this week, a phenomenon that's clearly reflected in the variability in performances.

Scores for the Head-to-Head Tipsters ranged from a low of 1 recorded by Ride Your Luck (which, evidently, didn't) to a high of 7 recorded by seven Tipsters, amongst them Bookie_3, Bookie_9, Bookie Knows Best and, I'm pleased to say, C-Marg.

Bookie_9's haul of seven sees it lead the MAFL Leaderboard, one tip clear of Combo_7 and one-and-a-half tips clear of Bookie_3. Home Sweet Home, despite snagging three correct tips this week, now finds itself in a familiar position at the foot of the table on just 20 from 36.

Variability also marked the performances of the Margin Predictors as mean absolute prediction errors (MAPEs) for the round spanned a range of almost 9 points per game from Combo_NN_1's 49.74 points per game to C-Marg's 58.57 points per game. In absolute terms as well, Predictor MAPEs were high this week, inflated by the 50-plus point victory margins in five of the contests.

Nonetheless, it's still quite close at the top of the Margin Predictor section of the Leaderboard as less than forty points separates Combo_NN_1 in 1st from H2H_Unadj_3 in 11th.

Four Predictors have offered margins so far this season that have placed them on the right side of the line market more than 50% of the time: Combo_7 and Bookie_3, who are near the top in terms of season MAPE, and Win_7 and C-Marg, who are both near the bottom. Once again this year we're seeing that Predictor performances in one market can be poor guides to performances in another.

C-Prob has surrendered top position amongst the Head-to-Head Probability Predictors, falling to third behind Bookie-OE and Bookie-LPSO, with Bookie-RE not far behind. And, looking at the final section of the Leaderboard, the Line Fund algorithm has maintained its positive probability score for the season to date.

SUPERMARGIN PERFORMANCES OF MARGIN PREDICTORS

Last week's sea of green in the table recording the Margin Predictors' SuperMargin wagering performances has given way this week to swathes of orange and red.

Had we wagered on each Predictor's prognostications in every contest so far this season, only for Combo_NN_1 would that have proven a profitable strategy. It has selected the correct bucket on 6 occasions, sufficient to record a +19.4% ROI.

Wagering instead only when a Predictor foresaw a home team win or a draw would have led to profit for three Predictors: Bookie Actual, Combo_7 and Combo_NN_1. The ROI would have been greatest had this strategy been applied to Combo_NN_1's predictions where its six winning wagers from 25 attempts would have yielded a +72% ROI.