2018 - Round 2 : Overs/Unders Update

/

Whenever MoSSBODS and MoSHBODS have opinions so significantly different from the bookmakers', my prior, I admit, tends always towards MoS being in error, but if I always acted based on that prior I'd never have the admittedly occasional windfall rounds.

Still, I'm nervous.

With that by way of unsubtle caveat, let me reveal that this week sees significantly lower team scores and game total forecasts from the MoS twins than from either of the TAB and Centrebet bookmakers.

Here's the detail.

From this we can draw the following:

HIGH-SCORING GAME

  • MoSSBODS and MoSHBODS : Lions v Dees
  • TAB and Centrebet : Kangaroos v St Kilda

LOW-SCORING GAME

  • MoSSBODS : Carlton v Gold Coast, or Sydney v Port Adelaide
  • MoSHBODS : Carlton v Gold Coast
  • TAB and Centrebet : Sydney v Port Adelaide

HIGH-SCORING TEAM

  • MoSSBODS : Melbourne
  • MoSHBODS : Essendon
  • TAB and Centrebet : St Kilda

LOW-SCORING TEAM

  • MoSSBODS and MoSHBODS : Port Adelaide
  • TAB and Centrebet : Port Adelaide

This week I've used simulations and MoSHBODS' team scoring opinions to estimate a probability for each team that it will record the round's highest or lowest score.

These are as shown in the table at right and are quite revealing in what they suggest about the difficulties of selecting the correct team in these markets. Ten teams, for example, have about a 1 in 20 or better chance of finishing as the round's high-scorer, and eight teams have about the same chance of finishing as the round's low scorer.

As well, if the estimates are broadly correct, no single team has better than about a 1 in 8 chance of finishing as high-scorer or as low-scorer, which isn't a huge reduction from the 1 in 18 chance they'd have if the outcome was completely random. And, even the team most likely to be high-scorer (Essendon), has about a 1 in 50 chance of being low-scorer, and the team most likely to be low-scorer (Port Adelaide) has about a 1 in 70 chance of being high-scorer.

It reminds me a little of Melbourne Cup fields ...

Now perhaps I've modelled too much variability in individual team scores, but I suspect that what we see here can't truly be described as a drastically misguided view.

Anyway, on to the week's wagers which, due to the significant differences of opinion between MoSSBODS and, in particular, the TAB, are numerous.

In seven games, in fact, MoSSBODS differs in its opinion from that of the TAB bookmaker by more than the requisite minimum 6 points, in each case with MoSSBODS forecasting a lower total. That has led to the seven unders wagers shown above.

PERFORMANCE TO DATE

Finally, let's take a look at the performance of the four forecasters in Round 1.

We see from this table that:

  • All four forecasters, on average, underestimated home team scores (by about 5-10 points per team) and overestimated away team scores (by about 6-9 points per team) and, consequently, underestimated margins from a home team perspective (by about 15-16 points per game).
  • MoSSBODS and MoSHBODS, on average, underestimated total scores (by about 1-2 points per game)
  • TAB and Centrebet, on average, overestimated total scores (by about 1.5-4.5 points per game)
  • TAB had the lowest mean absolute error (MAE) for game margins and for home team scores
  • Centrebet had the lowest MAE for game totals and for away team scores