2014 - Round 13 : Yet Another Dozen

There are weeks - take this one as a compelling example - where I partly wish I'd allowed myself veto rights for MatterOfStats wagers. Of course, it's exactly this kind of freedom that causes many gamblers, to their detriment, to overlay their intuitions and biases on their more objective assessments and analyses of form and ability.

That's exactly why I've not given myself that same freedom but, frankly, the prospect of 12 more SuperMargin wagers makes me cringe.

Round 13 starts off sedately enough with none of the three Funds fancying the Blues' chances even with 5-and-a-half goals start in their match up against the Hawks at the G on Friday night. Saturday brings our sole Head-to-Head wager, which this week is larger than it might otherwise be, the Head-to-Head Fund now liberated to the greatest extent it's afforded all season. It's used that liberty to splurge 4% on the equal-favourite Crows at $1.90.

Also on Saturday, we've three Line bets, one of those on the Crows, another on the underdog Tigers, and the third on the Swans giving Port Adelaide what seems to me to be a fairly generous 16.5 points start (but, see paragraph 1). Plus, we've the first half-dozen of our egg tray of SuperMargin wagers, none of this first lot, mercifully, including Chasms of Despair.

Then, on SuperMargin Sunday we've the remaining six Margin Fund selections - sporting two Chasms, one especially yawning - buttressed only by a Line bet on Geelong giving over 10 goals start to St Kilda. 

That all makes for another paper-clip strewn portfolio of profit and loss profiles for the nine games.

The Crows, Swans and Tigers offer the week's largest upside, each capable of lifting the Recommended Portfolio by at least 2.5c, whilst Freo, Port Adelaide, the Roos and the Saints collectively concentrate the majority of the round's downside, each imperilling 2.5c or more of Original Funds. Best case, the Portfolio rises by 15c, worst case it falls by 14c.

Of the three home teams carrying that maximum upside, only Sydney has been a net contributor the Recommended Portfolio's profitability so far this year, and theirs has been the smallest of the positive contributions. Port Adelaide's made the greatest contribution, tipping in over 7c on its own, ahead of five teams - Geelong, GWS, Hawthorn, the Kangaroos, and Melbourne - who've each contributed between 2c and 3c apiece. The two Coasts have been the largest destroyers of value, each costing Investors around 3.5c to 4c.

Melbourne's record when wagered on Head-to-Head is particularly unimpressive standing as it does at 0 from 5. In contrast, their record when wagered on in the Line market is a spectacularly impressive 4 from 5, only slightly less laudable than Port Adelaide's 5 from 6. St Kilda have the least attractive Line market record for Investors having never won on any of the four occasions on which they've carried our hopes.

Like Melbourne, however, the Saints' poor record in one market is counterbalanced by their solid performance in another. The Saints have a 2 and 2 record in the SuperMargin market, which has been enough to lift the value of the Margin Fund by almost 14c (and hence the Recommended Portfolio by almost 3c). The poorest records in the SuperMargin market belong to Fremantle and West Coast who have both racked up 0 and 8 performances, made only slightly better in Fremantle's case by the fact that those wagers have been made at a slightly smaller average bet size.

Looking, instead, at our wagering in terms of the teams we were wagering against - that is, the away teams in the contests we've bet on - Sydney has been most lucrative for us in the Head-to-Head market on account of a single profitable wager against them when they met GWS in Round 1. Port Adelaide has been least lucrative for us in this same market.

On Line betting, Fremantle and West Coast have both contributed 4.4c to the Line Fund, while Sydney has been responsible for handing 5.4c back to the TAB. GWS and Hawthorn head the list of teams in terms of returns from SuperMargin wagering, and Melbourne and St Kilda sit at the foot of that same list.


Four games have this week most troubled the Head-to-Head Predictors: Essendon v Melbourne, where eight have sided with the underdog Dees; Sydney v Port Adelaide, where nine are tipping an upset Port win; West Coast v Gold Coast, where 12 have selected the away from home and underdog Suns; and Adelaide v Kangaroos, where 13 (including the TAB) have picked the Roos and Bookie_3 is tipping a draw.

These relatively high levels of disagreement have, almost as a matter of course, led to high Disagreement Indices for each Tipster this week. Shadow has the round's highest Disagreement Index at 46%, and no Tipster has an Index under 20%. Overall, the average Index is 26%, meaning that two randomly chosen Tipsters in a randomly chosen game will differ in their selection about 1 time in 4.

The Margin Predictors are, again, far less split, and are on both sides of the outcome in only three games. In the West Coast v Gold Coast game, Win_3 and Win_7 foresee very narrow upsets, in the Adelaide v Kangaroos game the RSMP twins, Combo_NN twins and C_Marg are the only ones expecting the Roos to back up last weekend's extraordinary victory, and in the Essendon v Melbourne game it's the Win duo again and the H2H quartet that have opted for a narrow Dees win.

The first six games of the round all have small prediction ranges of about 20 points or less. Sunday's games have much wider ranges, the lowest being the 39-point range for the Pies v Dogs game and the highest the 55-point range for the Cats v Saints game.

Bookie_3 has the round's greatest Mean Absolute Deviation from the all-Predictor average and Combo_7 has the smallest. Win_7, however, is the most extreme Predictor in the greatest number of games, enjoying that status in four games, one more than both Combo_NN_1 and Combo_NN_2. C_Marg, in a rare display of conformity has managed to be the extreme Predictor in only two contests. 

The three contests that have kept the Margin Predictors from unanimity have done the same thing to the Head-to-Head Probability Predictors as WinPred, alone, has sided with the Suns over the Eagles; C_Prob, alone, has opted for the Roos over the Crows (with Bookie-RE and Bookie-OE abstaining); and WinPred and the two H2H Predictors have selected the Dees over the Dons.

C_Prob has assembled the round's most deviant - in a purely statistical sense - set of probability assessments, and WinPred has served up the second-most deviant. Both of them have made the most extreme assessments in five contests, more than any other Predictor. Together, the pair have also conspired to produce the largest range of probability assessments for any single game this week, C_Prob's assessment of Essendon's chances at 84% being some 39% points different from WinPred's 44% assessment.

The Line Fund algorithm continues to make what are sometimes referred to in probability forecasting lingo as "sharp" (ie nearer 0 and 1) probability assessments, including rating the Dogs as 84% chances, the Dees as 81% chances, the Swans as 79% chances, the Hawks as 72% chances, the Tigers as 68% chances, and the Crows as 65% chances.


ChiPS has again this week been heavily influenced in its margin predictions by the Ratings Difference of the competing teams. The correlation between Team Rating Difference and margin prediction this week comes in at +0.95.

In three of the games the team with the higher Rating has the poorer form as assessed by ChiPS: West Coast, Brisbane Lions and Geelong. Regardless, in all three cases, these teams are still tipped by ChiPS to win their respective contests. 

Based on the current TAB prices on offer, ChiPS recommends wagers on the Tigers at $4.60, the Eagles at $1.55, and the Dons at $1.36. It's nice that it's decided to wager on other than rank underdogs for a change.

C_Marg, as has been well-chronicled here on MatterOfStats, has lagged all other Margin Predictors this season and I've this week been wondering which teams have been the major cause of this lacklustre performance. 

Prior to performing the analysis I'd guessed that St Kilda were a significant contributor to C_Marg's high Mean Absolute Prediction Error, but I'd no idea that Sydney and Hawthorn had inflicted similar damage.

You can see the quantification of those detrimental impacts in the table at left, where you can also see that the Western Bulldogs, playing at home and playing away, have been especially well-predicted by C_Marg, as have Port Adelaide and Essendon but only when playing away, and Brisbane Lions, Carlton, Collingwood, Fremantle, Geelong and Melbourne but only when playing at home.

In a future blog I might perform a similar team-by-team analysis for other of the Margin Predictors and also attempt to tease out the causes of the relative inaccuracies in C_Marg's predictions for each team - are they due to inaccuracies in its assessment of Team Ratings, Team Form, Home Ground Advantage or Interstate Status?