2011 Round 9: Attritional Wagering

In this round, more than any before it this season, the terms "Home" and "Away" really mean something. For seven of the week's eight contests the designated Home team has at least 7 games more venue experience than their opponents, and in six of the games the Away team has played at the venue on no more than 3 occasions in the past year. Only Essendon has any right to feel that it's meeting its opponent on anything close to neutral territory and even then it has only half the venue experience of the Home team.

The Head-to-Head Fund's assessment of all this is that this home team advantage has been priced entirely appropriately in four of the contests, nearly so in two more, and not at all well in the remaining two. Thus it is that Investors find themselves with a 0.2% head-to-head wager on the Blues, a 0.9% wager on the Lions, a 4.0% wager on the Tigers, and a 4.7% wager on Port Adelaide. In aggregate, at just under 10% the proportion of the total Fund in play this week is by far the lowest this season.

The Tigers bet is an interesting one as there's clearly been some interest in them this week from punters other than us. Around noon on Wednesday, the usual time at which I insist that the Funds make their final deliberations before wagering, Richmond were trading at a price of $3. Earlier in the week, on Monday, the Tigers had opened at $3.30 but fell to $3 within about an hour of the markets being posted. The Head-to-Head Fund liked them at $3.00 and so, as I've just noted, took them for 4% of the Fund. Within about an hour of our wager (not that I'm inferring any causality) they shortened again to $2.80.

Meantime on the line market the Tigers opened +18.5 points at the standard $1.90 price on Wednesday at noon, to which the Line Fund said 'yes please' for 2% of the Fund. This price also fell within an hour of it being posted, in this case to $1.80. I'm never quite sure what to make of these sorts of market movements on underdogs, but I've no doubt we'll be reading about the "smart money" on the Tigers in next week's news should the Tigers run out winners - and nothing at all should the Tigers lose.

Our line bet on the Tigers is accompanied by 7 more line wagers, 6 of them on the Away teams. The two largest bets - for 4.3% and 4.5% of the Fund - are on Geelong and Melbourne in the first two games of the round, and the smallest wager is on the Roos for just 1.4% of the Fund. Combined, the eight wagers represent almost 25% of the Fund, which is roughly the proportion of this Fund that's been in play each week so far this season. At the current rate of activity we'll turn the funds in the Line Fund over between 6 and 7 times across the entire season. It'd be unfair to claim that we aren't giving the TAB Sportsbet bookmaker a good look at our money.

Here's the detail:

No single game can alter the value of Investor portfolios by more than 5% in either direction this round. A Richmond upset on Saturday will be the source of greatest joy to Investors, lifting portfolios by 4.9%, while greatest despair looms on the back end of a Freo victory by a margin of 1 to 12 points, which would knock 3.4% from portfolio values.

Other fiscally important and joyous results would be a Port Adelaide victory (+2.5%), a Melbourne win or narrow loss (+2.0%), or a Geelong victory by 15 points or more (+1.9%). Significant portfolio losses would stem from a Saints win by15 points or more (-2.2%), a Geelong win by 1 to 14 points (-2.2%), or an Eagles win by 4 points or more (-1.9%).

Here's the full set of possibilities (with a reminder that all images in MAFL are clickable):

Majority head-to-head tipster support lies with the favourites in six of this round's contests, the exceptions being the Saints v Dees clash where the tipsters favour the Dees 8-5, and the Sydney v Hawthorn matchup where the tipsters narrowly favour the Hawks 7-6. In every contest there are at least 3 dissenting tipsters.

Three matches are most important for separating the leading heuristic-based head-to-head tipsters, BKB, Ride Your Luck (RYL) and Short-Term Memory II (STM II). In the Saints v Dees clash BKB is on the Saints whereas RYL and STM II are on the Dees; in the Sydney v Hawthorn game it's BKB again on its own, in this game tipping the Swans while RYL and STM II favour the Hawks; and in the Eagles v Dogs contest it's STM II that's the odd tipster out in selecting the Eagles where BKB and RYL favour the Dogs.

The Margin Predictors are also majority-supporting six favourites, though they're supporting the favourites in the two contests where the Head-to-Head tipsters are supporting the underdogs, and supporting the underdogs in the Port Adelaide v Fremantle, and Lions v Roos games.

With the exception of the Saints v Dees, and the Pies v Crows games, for both of which the Margin Predictors are unanimous, as we saw for the Head-to-Head tipsters there is considerable disagreement amongst the Margin Predictors. For four of the contests the average predicted margin of victory is under 2.5 points, and in three more the average is under 6.5 points. Only the Collingwood v Adelaide game is universally predicted to be a comfortable win for the favourites.

This week we'll look at the manhattan distances between the Margin Predictors slightly differently, firstly by reporting this week's pairwise distances on a per game rather than an aggregate basis. 

We can see from this graphic how starkly different are Combo_NN_1's margin predictions from most other Margin Predictors'. For example, the average absolute difference in its margin predictions compared with those of WinPred_3 is over 19 points per game. 

Bookie_3, currently the Predictor with the best MAPE, is most different from the two ProPreds. More importantly, it's about 9 points per game different from Combo_NN_1, 4 points per game different from Combo_NN_2, and 3 points per game different from Combo_7, which are collectively the Predictors filling spots 2 through 4 on the MAPE ladder.

So, how different are these differences from those that we've seen across the 9 rounds of the season? This next graphic takes the difference data from the graphic above and recalculates it relative to the average difference in the margin predictions for each pair of predictors across the 69 games they've made predictions for this season.

Positive numbers represent larger differences in the average difference for Round 9 compared to the average difference across the season and negative numbers represent the opposite situation. For example, the -0.8 shown for Bookie_9 and Bookie_3 means that the average difference in the Round 9 margin predictions for these two Predictors, which is 5.8 points per game, is 0.8 points lower than the all-season average (which we can infer must be 6.6 points per game).

The slew of negative numbers in the first column tell us that the margin predictions of virtually every Predictor are this week closer to Bookie_3 than they typically have been across the entire season. Given the strong MAPE performance of Bookie_3, that's probably not a bad thing for these other Predictors to be doing.

Comparing the probability assessments of ProPred, WinPred and H2H (Unadj) with those of the TAB Sportsbet bookmaker this week we find significant disagreement from ProPred and WinPred in the Carlton v Geelong, Port Adelaide v Fremantle, Lions v Kangaroos, and Richmond v Essendon games. In each case ProPred and WinPred believe that the Home team underdogs should instead be favourites.

H2H (Unadj) broadly agrees with ProPred's and WinPred's assessments of the Port v Freo and Lions v Roos games, and also disagrees with the TAB Sportsbet bookmaker in the Sydney v Hawthorn, and the Eagles v Dogs games where it favours the Away team underdogs.