2018 - Round 12 : Overs/Unders Update

/

Seven games this week, and not much to make you happy if you're craving attacking, high-scoring football, with only the Lions v Dons game expected to breach the 180 mark for total score.

That game aside, if we're to believe the bookmakers we're looking at one game with a total under 150, another with a total under 160, three more with a total under 170, and the Dees v Pies game with an expected total just under 180. 

The MoS twins are more optimistic about four of the games, but less optimistic about three and therefore, on balance, slightly more optimistic overall. They're forecasting an average total score for the round of about 167 or 168 points, just 1 or 2 points average the season average to date.

Even with their varying levels of optimism about the likelihood of high scores this week, for the first time (I think) this season all four forecasters agree about the likely high- and low-scoring teams and games.

MOST LIKELY HIGH-SCORING GAME

  • All : Brisbane Lions v Essendon (180 to 188.5)

MOST LIKELY LOW-SCORING GAME

  • All : Port Adelaide v Richmond (140.5 to 157)

MOST LIKELY HIGH-SCORING TEAM

  • All : GWS (103 to 112)

MOST LIKELY LOW-SCORING TEAM

  • All : Gold Coast (60 to 60.5)

Last weekend, I analysed MoSHBODS' performance to the end of Round 10 in estimating which would be the low- and high-scoring teams and games. I did this by comparing the log probability scores of its probability estimates with what could have been achieved through naive forecasting (ie by assigning equal probabilities to all teams and games).

Using that metric, I found that MoSHBODS had performed better than a naive forecaster in assigning probabilities for the high- and low-scoring teams (doing better on the former), but slightly worse than a naive forecaster in doing the same for the high- and low-scoring games.

I might write the methodology and analysis for this up if I get some time.

MoSHBODS' probability estimates this week have GWS as clear favourites for the round's high-score, ahead of Melbourne, Essendon and the Brisbane Lions, who all have roughly equal estimated probabilities. GWS' opponents, Gold Coast, are very firm favourites to be the round's low-scoring team.

The Lions v Dons game is estimated as being most-likely to be the round's high-scoring game, with an estimated probability of almost 26%, while the Saints v Swans, and Power v Tigers games have roughly equal estimated chances of finishing as the round's low-scoring game, both with estimated chances of around 21 to 22%.

WAGERS

Investors face wagers in six of the seven contests this week, avoiding only the Geelong v North Melbourne game where the overlay relative to the TAB total of 162.5 fell (effectively) just one point short of the necessary 6 points.

There are three unders wagers, with overlays ranging from 7 to 9 points, and three overs wagers with overlays ranging from 8 to 14 points in games where the forecasts are currently for rain, showers or even storms.

PERFORMANCE TO DATE

Last week was a below-average one for MoSSBODS and MoSHBODS in terms of the totals markets, with MoSSBODS on the correct side of the TAB's totals in four of the nine games, and Centrebet's totals in just three. MoSHBODS did only slightly better, landing on the correct side of both the TAB's and Centrebet's totals in four of the nine games.

That gives MoSSBODS a season-long 52% record against the TAB and Centrebet, and gives MoSHBODS a 51% record against the TAB and a 54% record against Centrebet.

MoSSBODS' performance in games on which it has wagered is summarised in the table at right, which this week includes an overs/unders breakdown for the first time this season and which reveals a marked difference in MoSSBODS' success when taking overs versus unders. 

Of the 15 times it's opted for an overs bet, MoSSBODS has been right only 33% of the time. In contrast, in the 37 games where it's taken the unders it has a 65% record. Obviously there's much greater variability in the overs success rate figure because it is based on a much smaller sample, but it is a number we should continue to watch, especially since it is the only one this is negatively affected by MoSSBODS' weather-blindness.

Despite MoSSBODS slight bias for being on the wrong side of bookmaker totals last weekend, it nevertheless performed outstandingly on the mean absolute error (MAE) metric, recording the lowest of all four forecasters for game margins, home team scores, away team scores, and totals. 

Centrebet still leads, however, on the season-long view for home team scores, as does MoSHBODS on game totals and away team scores. The only change for the week was that the TAB bookmaker took top spot on game margins from Centrebet.