2011 Round 2 : Regress and Regret or Consolidate and Confirm?

Often we make implicit decisions even when we kid ourselves that we're not making a decision at all - we're just letting "nature take its course". For example, I heard the other day about the Principle of Indifference, which states that, in the absence of evidence to the contrary, all possible outcomes of an event should be treated as equally likely. This sounds to me like an invitation to be fleeced by somebody who has greater knowledge about an event than you do and who is willing to frame a market to take advantage of this - a situation that is sometimes referred to an information asymmetry.

The decision not to change from an otherwise predestined course of action is a decision not to change from that predetermined course. And the decision not to change the default initial rating for a footy team is a decision to apply the default initial rating to that team.

Forced to rate the Gold Coast amongst the current crop of AFL teams, where would you rate them? Below all other teams? Below half of them? Somewhere else? I don't have a good answer to that question either, so I've allowed the MARS Rating algorithm to rate the Gold Coast at 1,000 MARS Rating points, the same rating that all other teams started with as at the first game of season 1999. As a direct consequence of this decision, the Gold Coast appear to the Head-to-Head and Line Fund algorithms as an average team playing at home and not as a newly formed team playing its inaugural match in the competition.

So, the Gold Coast appears as the subject in one of only two head-to-head wagers this week, and as the target in one of six line wagers. Here's the detail:

(It's interesting to note that the TAB Sportsbet bookie has persisted with starts of less than 6.5 points this week. We'll need to keep an eye on how the Line Fund performs in games with sub-goal starts as it's seen no such games in the data from previous seasons on which it has - predominantly at the moment there having been few rounds in the current season - been trained.)

A Gold Coast victory offers the largest reward to Investors this week and even a draw would be a welcomed result, surpassed only by a 27 or more point victory by the Saints.

Here's what all possible outcomes would mean for Investors: 

Seeing this, it's apparent that a St Kilda loss to the Tigers on Friday night would spoil the remainder of the weekend, at least from a wagering point of view.

Whereas, in our wagering, the default course of action has had obvious and (in hindsight) predictable consequences, it was much harder to predict that my decision to treat draws as losses in the revamped version of the Heuristic algorithms would have such an early and dramatic effect.

Had Sydney's draw with Melbourne been treated as a win, many of the Heuristics would have selected the Swans this week rather than the Dons. Instead, all but BKB and HSH have opted for Essendon.

(There's one other important change to this year's heuristics that will be of most significance early in the season and that is the determination of which team is higher on the ladder when only some of them have had byes. For the purposes of those heuristics that use ladder position, most notably CTL, such position is based firstly on the win percentage of each team - not the competition points accumulated - and then on For and Against Percentage.) 

You'll note that there's relatively little contention this week amongst the head-to-head tipsters about any of the games. Only in the Bulldogs v Lions, and Hawks v Dees matchups can we find as many as four dissenters from the mainstream view. What's maybe a little surprising though is that in only four of the games is the majority tipster pick the favourite.

In contrast, the majority Margin Predictor pick is the favourite in six of the eight games.

  • St Kilda v Richmond - unanimous support for the home team favourites, with predicted victory margins ranging from 23 to 62 points. Only 1 tipsters is not predicting that the Saints will cover the 26.5 point spread.
  • Kangaroos v Collingwood - the Pies are tipped by 9 of the 13 tipsters, by margins ranging from under a point to 63 points. The four tipsters that are predicting a Roos upset are suggesting victories by 3 points or fewer. Only two tipsters have Collingwood covering the 37.5 points spread.
  • Port Adelaide v West Coast - the margin tipsters are torn on this game and are split 7-6 in favour of Port. Predictions range from a Port Adelaide victory by 16 points to a West Coast victory by 11 points. Only one tipster foresees Port covering the 13.5 points spread.
  • Gold Coast v Carlton - three tipsters envisage a Gold Coast upset, by margins ranging from less than a point to 7 points. The remainder predict a Blues victory by anywhere between 1 and 55 points. Only two tipsters predict that the Blues will win by enough to cover teh 26.5 point spread.
  • Fremantle v Geelong - only three tipsters reckon that the Cats will prevail, and they each think it'll onlybe by 2 or 3 points. The remaining 10 tipsters believe that Freo will win, but by no more than 7 points, and only four of them have the Dockers covering their 5.5 point spread.
  • Dogs v Brisbane - the second game that's produced unanimity amongst the tipsters, this time for the Dogs who are expected to win by between 25 and 58 points. Only five tipsters predict that the Dogs will cover their 37.5 point spread.
  • Sydney v Essendon - here the Dons are favoured 8 to 5 and are expected to win by between 2 and 7 points. Those foreseeing a Swans victory do so by margins of between 4 and 26 points, this latter prediction being very much the outlier and coming from the determined-to-be-different neural network #1. Only three tipsters predict that the Swans wil cover their 5.5 points spread.
  • Hawthorn v Melbourne - it's either a Hawks victory or a 1 point win by the Dees according to the Margin Tipsters. Just three of the tipsters feel that the 16.5 points start for Melbourne wont be sufficient to get them home on line betting.

Lastly, let's look at the probability predictions of the Probability Tipsters.

You'll notice that ProPred, WinPred and H2H Unadjusted have the Roos as around even-money chances in their matchup with the Pies, which is significantly different from how the TAB Sportsbet bookmaker rates their chances. The only reason that Investors are not facing a large Roos head-to-head wager this week is because, at $5.65, they exceed the maximum threshold at which the Head-to-Head Fund is permitted to wager. This game will be an interesting initial test of the wisdom of that constraint. 

The only other aspect of this table that I'll mention is how much the Line Fund probability predictions changed when it transpired that Freo and Sydney were only giving 5.5 points start and not the 7.5 points start that I was expecting in each game and, on the basis of which, had performed earlier runs of the algorithms. With Freo giving 7.5 points start the Line Fund would have been on the Cats with about a 5% wager (rather than the 0.6% we now have on Freo), and, if the Swans were giving 7.5 points start, the Line Fund would have had about a 3% wager on the Dons instead of no bet at all on either team. I'm not sure whether to be impressed or alarmed by this level of sensitivity, but I'll have a better view of this too come Sunday.