2015 - Round 17 : Anything But Backward

The Australian Coat of Arms bears images of both a Kangaroo and an Emu, these animals having been chosen for the task, it's said, because they're both unable to take a backward step. Were there to be a MoS Coat of Arms (now there's an off-season task for someone more artistically talented than I) it would surely bear images of the Head-to-Head and Line Funds.

Both ended last week with their metaphorical tails between their imaginary legs, each down 27% on the season, yet they've returned this week to wager an amount in total larger than any they've wagered in a single round this season - a whopping 17.6% of the Fund.

The Head-to-Head Fund has outlaid 21.5% of the Funds with which it was originally entrusted on six home teams ranging in price from $1.60 to $14.00, with larger wagers, fortunately, on shorter-priced favourites, and smaller wagers on longshots. Because of this skewing of larger bets at smaler prices, and vice versa, all up the outlay-weighted average price for the wagers is just $2.10.

Those six head-to-head wagers are complemented by three line wagers totalling 15% of that Fund, two on the same teams that the Head-to-Head Fund fancies, and the third Investor's only bet on the Dogs this week.

GWS, one of the teams carrying both a head-to-head and a line bet, also carries the week's highest upside and downside, a 2-point or larger win worth over 5c to Investors, and a loss of any size costing 6c. Carlton, by sheer dint of the price we've secured ($14), carries next most upside, while the Lions, the other team attracting twin-Fund attention, carries next most downside. 

Just two games carry no wagers this week. Favourable results in the other seven would add almost 20c to the Overall Portfolio while unfavourable results would slice about another 18c from it.


With unanimity about the chances of two teams this week (Adelaide and West Coast), and near unanimity about the prospects of three more (Hawthorn, GWS and the Kangaroos), disagreement levels are far lower amongst the Head-to-Head Tipsters this week than last.

In the four other contests, minority groups contain anywhere from 6 to 10 Tipsters, the largest forming in support of the proposition that Freo will beat Richmond, and that the Dogs will beat the Pies.

Short Term Memory II is the most cavalier Tipster this week, though Home Sweet Home and Easily Impressed II aren't all that far behind. Neither BKB not Bookie_3, two of the Top 4 Tipsters on MoS, have the lowest Disagreement Indexes because both have opted for minority opinions in the Melbourne v St Kilda game. BKB has selected St Kilda because it's obliged to mirror Consult The Ladder when there are equal favourites in a contest, and Bookie_3 has predicted a drawn game. 


The all-Margin-Predictor Mean Absolute Deviation (MAD) this week has come in at 5.8 points per game per Predictor, the highest it's been since Round 11.

C_Marg and Combo_NN2 are responsible for a hefty slice of that mean deviation, they each sporting MADs of over 12 points per game. The Predictors nearest C_Marg on the MoS Leaderboard, RSMP_Weighted, Bookie_9, and RSMP_Simple, all have much lower MADs, in particular Bookie_9, which has the low MAD of the round at just 2.3 points per game.

Driving the average MAD up when viewed from a game-by-game perspective is the Carlton v Hawthorn game where the overall MAD is 14.7 points per Predictor and the range of predicted margins spans more than 8 goals from a 22-point Hawks win predicted by Win_3, to a 70-point Hawks win predicted by C_Marg.

Three other games have MADs greater than a goal: GWS v Geelong (7.3 points per Predictor), Adelaide v Gold Coast (6.8), and Melbourne v St Kilda (6.3). The Dogs v Pies game has the smallest MAD of just 2.4 points per Predictor.


This week's all-Head-to-Head-Predictor average MAD of 5.7% points per Predictor per game is relatively high in the context of the season, but low compared to last week's 6.6% points.

C_Prob has the highest MAD (10.8%) for the fourth successive week and for the tenth time this season, while Pro_Pred has the next-highest MAD (7.7%). Bookie-LPSO has the week's lowest MAD, slightly ahead of the two Head-to-Head Predictors, Bookie-RE, and Bookie-OE.

The GWS v Geelong, and Melbourne v St Kilda games have the highest game MADs, these MADs both most heavily influenced by the opinions of C_Prob and ProPred. In the first of those games, C_Prob's 43% assessment is at one end and ProPred's 76% at the other, while in the second we find C_Prob's 46% at the South Pole and ProPred's 78% at the North Pole. The Lions v Roos game has the smallest MAD of just 3.1% and an amazingly small prediction range of only 10% points from 22% to 32%.

Only one team is this week rated as better than a 60% chance of winning on line betting by the Line Fund algorithm: GWS, which the algorithm assesses as a 61% prospect. Five other teams are assessed as having chances of 56% or higher, however: Brisbane Lions (58%), Western Bulldogs (57%), Hawthorn (58%), Fremantle (56%), and Port Adelaide (56%).


The latest analysis reveals that the round-by-round correlations between:

  • All-Head-to-Head-Tipster Disagreement Indexes and Average Accuracy is now -0.37
  • All-Margin-Predictor MADs and MAE is now +0.23
  • All-Head-to-Head-Probability-Predictor MADs and Average Probability Scores is now -0.42

Broadly speaking then:

  • lower levels of disagreement amongst the Head-to-Head Tipsters have been associated with more accurate overall tipping
  • lower levels of disagreement amongst the Margin Predictors have been associated with more precise overall margin prediction
  • lower levels of disagreement amongst the Head-to-Head Probability Predictors have been associated with better-calibrated probability estimation