2016 Round 9 : Overs/Unders Update

It's a bundle of five Overs/Unders wagers this week, four on the Under and one on the Over, which makes the full weekend's festivities look like the following.

In total, the 13 wagers across the three Funds represent just under 9% of the original Overall Portfolio, making this, by amount wagered, the third most active round for MoS Funds this season, behind only the bet-fests of Rounds 2 and 4.

Here's hoping we can be somewhere even close to the ROIs of those rounds.

The week's Unders bets come in games where the TAB Total is somewhere in the 180.5 to 192.5 range, and the sole Overs bet comes in the Fremantle v Richmond game where the TAB Total is a season-low 155.5 points, eclipsing the 165.5 point Total offered for the Round 3 game between the Roos and the Dees at Bellerive. That game finished as a 136-131 win for the Roos, making it a very comfortable Overs result for anyone who got onto that.

MoSSBODS continues to err on the conservative side in terms of team scores, though this week it forecasts lower scores than the TAB for only 13 of the 18 teams. Six of those 13 teams are forecast to score more than two goals less by MoSSBODS than the TAB, however:

  • GWS: MoSSBODS forecast 87 points; TAB forecast 105 points; Difference 18 points
  • Western Bulldogs: MoSSBODS forecast 69 points; TAB forecast 87 points; Difference 18 points
  • Adelaide: MoSSBODS forecast 115 points; TAB forecast 131 points; Difference 16 points
  • Melbourne: MoSSBODS forecast 104 points; TAB forecast 118 points; Difference 14 points
  • West Coast: MoSSBODS forecast 95 points; TAB forecast 107.5 points; Difference 12.5 points
  • St Kilda: MoSSBODS forecast 94 points; TAB forecast 106.5 points; Difference 12.5 points

For two teams, Fremantle and the Gold Coast, MoSSBODS' forecast score is two goals go more greater than the TAB's.

Conservative team score forecasts result in conservative Totals forecasts, and MoSSBODS finds itself forecasting lower Totals in eight of the nine games this week, most notably in the four games in which we've Unders bets where MoSSBODS' Total forecast is anywhere between about 3 and 6 goals less than the TAB's.

In the one game where MoSSBODS' Total forecast is higher, the difference is only about 2 goals.

Despite its general conservatism throughout the season so far, MoSSBODS performance relative to the TAB across the first eight rounds remains encouraging. The various comparative mean absolute errors (MAEs) are now:

  • For Home Team Scores: MoSSBODS 22.3 points per team; TAB 19.9 points per team
  • For Away Team Scores: MoSSBODS 20.2 points per team; TAB 21.2 points per team
  • For Either Score: MoSSBODS 21.3 points per team; TAB 20.6 points per team
  • For Total Score: MoSSBODS 32.2 points per game; TAB 29.1 points per game
  • For Margins: MoSSBODS 29.6 points per game; TAB 31.0 points per game

Home Team Scores and Totals remain the two areas of MoSSBODS underperformance, Away Team Scores and Margins those of relative excellence.

A team-by-team analysis of actual versus expected scoring is informative:

In the table below we look at how each team has scored at home and away, in actual terms and relative to MoSSBODS' expectations.

So, for example, the first row reveals that, at home, GWS has generated 35.5 Scoring Shots (SS) per game against a MoSSBODS expectation of just 24.9 SS per game. They've converted those SS at a rate (57.0%) sufficient to average 136.8 points per game, which compares with the MoSSBODS expectation of 91.2 points per game, based on an assumed 53.3% conversion rate (the average rate of all teams in 2015).

Playing away, GWS have generated 2.6 more SS per game than expected, and scored 7.8 points per game more than expected. Overall then, GWS have generated 6.6 more SS per game and 26.6 points per game more than expected. That's the largest positive differences in terms of SS per game, and the second-largest in terms of points per game of all the teams.

As we scan down the Combined data we see that every team but one, Essendon, has generated more Scoring Shots per game than MoSSBODS has expected, though for most teams the excess has been less than 2 SS per game. How these excesses have converted into differences in actual scores relative to expectations has depended on a team's average Scoring Shot conversion rate.

Melbourne, for example, which so far this season has converted at an absurdly high 63.9%, has turned a 3.7 SS per game excess over expectations into a 27 points per game excess over expectations, while Fremantle, who've converted at an appalling 44%, have turned a 1.4 SS per game excess into a 5.4 points per game deficit. 

Whilst team conversion rates have, individually, varied quite widely, the all-team average of 53.6% is very close to what MoSSBODS expected. As such, at the aggregate level at least, we can ignore the impact of conversion rate on the differences between actual and expected scores, and focus on that which MoSSBODS was designed for: forecasting expected Scoring Shots.

With that in mind, if you scan the very last row of the table you'll see that, for home teams, MoSSBODS has, on average, forecast 2.7 SS too few per game and, for away teams, only 0.9 SS too few per game. That's why MoSSBODS has done so much better relative to the TAB on forecasting away scores than on forecasting home scores.