2012 Finals Week 4 - Wagers & Tips : We're Cheering For a Margin, Not a Team

I wish it were different. The Line Fund's disinterested, the Head-to-Head Fund's prohibited, while the Margin Fund's oblivious to its long losing streak. So it is that Investors find themselves with a single wager on the Grand Final: the Hawks to win by 30 to 39 points for 2.5% of the Margin Fund at $8. The difference between the success and failure of that single bet is 4c, the upside a 3.5c gain and the downside an all-too-familiar 0.5c loss.

Amongst MAFL Tipsters and Probability Predictors, a Hawks win is the unanimous opinion, as indeed it is amongst the Margin Predictors, but none other than Combo_NN_2 amongst them foresee the margin being as large as 30 points.

The Head-to-Head based Predictors are tipping the narrowest margins, reckoning that the Hawks will get home by 10 points or fewer, while five more Predictors assess the Hawks' advantage at 2 goals or less. (The TAB is offering Sydney only 13.5 points start on line betting - though at the moment he's also offering slightly better than even money odds for Sydney along with that handicap. With the Hawks' head-to-head price of $1.45 I'd assumed that they'd be giving about 18.5 points start, so I can see why the money might be with the Hawks given such a small handicap. The Line Fund, of course, thinks otherwise.)

With the Line Fund yet again shunning the Hawks on the line market and this week assessing them as only 28% chances of covering the spread, I got to wondering about the Line Fund algorithm's performance in relation to the Hawks so far this season. The table that follows shows, for every team, its winning percentage on line betting this year at home and away, the average probability assessment of the Line Fund algorithm when that team was playing at home or away, and the average probability score that the Line Fund algorithm recorded for that team on the basis of these probability assessments and the actual line betting results.

(Note that the probability score for a probability estimate of p made about a team that won on line betting is 1+log(p) and the probability score for the same probability estimate about a team that lost is 1+log(1-p). In both cases the logs are calculated base 2, so the score is measured in bits.)

From this table we can see, for example, that Hawthorn have won on line betting when at Home (as they are deemed to be this week) only 54% of the time and that the Line Fund algorithm's average probability assessment has been 46%. Nonetheless, it's managed to generate an average probability score of +0.07 bits from its Hawthorn-at-home probability estimates. On that basis I suppose it's hard to be too unhappy that the Line Fund has chosen not to wager on the Hawks this week, though I'd still question the 28% figure.

Looking more broadly than at just the Hawks' data we find that: 

  • Home teams have yet again registered a better than 50% win-loss record. Someone, somewhere, more patient and diligent than I, must surely be cranking out profits from this fact.
  • The levels of variability in line betting performances across teams is very high. The Eagles won three-quarters of the time on line betting when at home but only one-third of the time when away, whilst the Swans won two-thirds of the time at home and away, and GWS won over 80% of the time at home and less than 30% of the time when away.
  • The Line Fund algorithm was unable to produce a positive probability score for any team both at home and away, the closest probably being the Suns for which its probability scores averaged -0.01 when the Suns were at home and +0.08 when they were away.
  • Adelaide, Carlton, Port Adelaide and Melbourne have been responsible for the Line Fund algorithm's worst combined home and away records this season.
  • On the positive side, the Line Fund algorithm's average probability assessment of home teams was about 53%, and home teams' actual success rate was also about 53%. On average then, the Line Fund algorithm was extremely well-calibrated.