2017 - Round 6 : By How Much?

/

This week, the forecasters - well, at least those that don't take bookmaker input - are split more in terms of victory margins than they are in terms of picking winners.

Amongst the Head-to-Head Tipsters, only one game has produced minority support that numbers more than two. That's come in the Dons v Dees game where six of the nine Tipsters have selected the (narrow) underdog Dees to prevail.

Beyond that game we find just two Tipsters predicting an upset Saints win, two an upset Blues win, one an upset Lions win, and one an upset Suns win.

All up that's left us with an all-Tipster Disagreement Index of just 19%, equal to that we had last week and higher only than the 14% figure in Round 2.

There's more debate, however, about the margins of victory, with C_Marg and the two MoS' in particular having more radical opinions than their peers. C_Marg has opted for the most extreme home team victories in six of the games, while MoSSBODS_Marg has been the most bearish of all the Predictors about home team chances in four of the contests. MoSHBODS_Marg has avoided being the most extreme Predictor in any game but has still produced a mean absolute deviation of 5.4 points per game.

Those three aside, there are unusually high levels of agreement otherwise amongst the remaining seven Predictors.

In the end, the mixture of minority dissent and majority accord has produced an all-Predictor average MAD that's about equal to the all-season average of 4.8 points per Predictor per game.

The two clashes that have caused most disagreement are the Carlton v Sydney, and Brisbane Lions v Port Adelaide matchups where the MADs have come in at 8.4 and 7.0 points per Predictor respectively and where, in the first of those games, the forecasts have spanned almost a 7-goal range.

Amongst the Head-to-Head Probability Predictors, it's C_Prob and MoSSBODS_Prob that stand out, C_Prob finding itself as the most extreme forecaster in seven games, and MoSSBODS_Prob in six.

From a game-by-game perspective, here it's only the Blues v Swans game that's produced an especially large MAD, driven up by Bookie-RE's 14% estimate and MoSSBODS_Prob's 47% estimate, with the result that the all-Predictor MAD for the round is the second-lowest of the season.

WAGERS

The lowish MAD of MoSHBODS_Prob and the highish MAD of MoSSBODS_Marg have had predictable effects on this week's wagering: just three head-to-head but five line wagers.

It's the twin wagers on Carlton though that will almost entirely determine how fondly or otherwise the upcoming weekend will be remembered since the difference between the best and worst outcomes in that game span almost 11c.

In the remaining eight games, only the Lions v Power game represents a swing greater than 1c.

The line wager on Melbourne, though small, is an interesting one because in making it I had to decide which of these two propositions was to be preferred:

  • Melbourne +1.5 points at $1.87 (Centrebet)
  • Melbourne +2.5 points at $1.80 (TAB)

To arrive at a decision on this we need to make assumptions about the statistical distribution from which game margins are drawn. In the past I've found that a Normal distribution with a mean equal to the expected margin and a standard deviation of about 34-38 points provides a reasonable empirical approximation for this purpose (for example, see this post or this later post).

Assuming that MoSSBODS_Marg's estimate of the true expected margin is correct (ie a 6.4 point Melbourne win) and using a standard deviation of 36 points, we can calculate the two probabilities that we need using Excel as follows:

  • Prob(Melbourne loses by 1.5 points or less) = NORMDIST(1.5, -6.4, 36, TRUE) = 58.7%
  • Prob(Melbourne loses by 2.5 points or less) = NORMDIST(2.5, -6.4, 36, TRUE) = 59.8%

Given those probabilities we can now calculate the expected return of each bet:

  • At $1.87, the first bet has an expected return of 1.87 x 58.7% - 1 = +9.7%
  • At $1.80, the second bet has an expected return of 1.8 x 59.8% - 1 = +7.6%

So, the Centrebet proposition is to be preferred given these assumptions.

Even if we use, instead, an expected margin of 0.5 points in favour of Essendon - a number more in keeping with the bookmakers' opinions - we still find that the Centrebet proposition is to be preferred (albeit that it's now a loss-making one, on average).

MoSSBODS AND MoSHBODS TEAM SCORING OPINIONS

I'm not yet ready to assume that the TAB and Centrebet will release all of their Totals markets in the next 24 hours, so I'm again going to provide their opinions about team scores in advance of making any bets.

(Note that it's only MoSSBODS' opinions about totals that matters in terms of what the Over/Under Fund will do. I'm providing MoSHBODS' opinions too only for comparative purposes. I'll save any discussion of this comparison for a subsequent post once the Totals markets are available.)