2018 - Round 14 : Overs/Unders Update

/

The bookmakers expect the Lions v Giants game to elevate average scores most this week, but even with that game slated to produce 185 points, they still have an all-game average of only about 167 to 168 points per game, which is only a point or two above the 2018 average.

The MoS twins have considerably lower expectations for the Lions' contest, but higher expectations elsewhere, to the extent that their overall expected average total is roughly the same as the bookmakers'.

Those of you who don't follow me on Twitter (and, doubtless, many of you who do) won't have seen the chart below, which decomposes 2018 scoring into its scoring shot and conversion rate components.

It shows that scoring this year has been lower than in recent seasons both because of fewer scoring shots being generated, on average, by teams in each game, and because those scoring shots have been converted into goals at a lower rate (ie have produced a lower mean score per scoring shot). We currently have 1929 level scoring shot generation, and 1982 level scoring shot conversion.

Returning to the latest set of forecasts, we find relatively high levels of disagreement this week about the likely high- and low-scoring games and teams.

MOST LIKELY HIGH-SCORING GAME

  • MoSSBODS : Brisbane Lions v GWS, and Western Bulldogs v North Melbourne (172)
  • MoSHBODS : Brisbane Lions v GWS, and West Coast v Essendon (170)
  • TAB & Centrebet : Brisbane Lions v GWS (184.5 to 185.5)

MOST LIKELY LOW-SCORING GAME

  • MoSSBODS & MoSHBODS : Hawthorn v Gold Coast (157)
  • TAB & Centrebet : Collingwood v Carlton (156.5 to 157.5)

MOST LIKELY HIGH-SCORING TEAM

  • MoSSBODS : Hawthorn and Collingwood (104)
  • MoSHBODS : Collingwood (104)
  • TAB & Centrebet : Hawthorn (107.5 to 109)

MOST LIKELY LOW-SCORING TEAM

  • MoSSBODS & MoSHBODS : Gold Coast (52 to 54)
  • TAB & Centrebet : Carlton (55 to 55.5)

MoSHBODS did poorly in forecasting the extremes of team and game scoring last weekend, attaching:

  • 14.8% probability to the highest-scoring team, Port Adelaide (against a naive forecast of 8.5% since 12 teams were playing)
  • 7.7% probability to the lowest-scoring team, Adelaide (against a naive forecast of 8.5%)
  • 8.3% probability to the highest-scoring game, Port Adelaide v Western Bulldogs (against a naive forecast of 16.7% since 6 games were played)
  • 13.7% probability to the lowest-scoring game, Hawthorn v Adelaide (against a naive forecast of 16.7%)

Over the course of the season it's still done worse than a naive forecaster in estimating probabilities for the high- and low-scoring games, and better than a naive forecaster in estimating probabilities for the high- and low-scoring teams.

This week, MoSHBODS has Collingwood as, narrowly, the most likely high-scoring team, and the Gold Coast (by some margin) as the most likely low-scoring team.

It'll do better than a naive forecaster as long as one of Collingwood, Hawthorn, West Coast, GWS, North Melbourne or Melbourne finish as high-scorers, and if any of Gold Coast, Carlton, or Essendon finish as low-scorers.

The Lions v Giants, and Power v Dees games are estimated as being equally most-likely to be the round's high-scoring game, with an estimated probability of 19%, while the Hawthorn v Gold Coast game is estimated as having the highest probability of being the round's low-scoring game, with an estimated probability of almost 26%.

MoSHBODS will do better than a naive forecaster as long as one of the Lions v Giants, Power v Dees, Eagles v Bombers, or Dogs v Roos games finish as the high-scorers, and if either the Hawks v Suns, or Pies v Blues games finish as the low-scorers.

WAGERS

Investors face four wagers this week, two overs and two unders.

Estimated overlays across the four bets range from about 1 to 2 goals, though the forecasts for both Thursday night's and Sunday afternoon's game, where we've overs bets, are for possible showers, which would make those overlay estimates a little generous. 

Rain is also forecast for the Saturday game in Tasmania, however, where we've an unders bet, so there's that.

PERFORMANCE TO DATE

MoSSBODS finished on the correct side of the TAB's and Centrebet's totals in three of the six games last weekend, and MoSHBODS in only two.

That maintained MoSSBODS' season-long 50% record against the TAB and Centrebet, and gives MoSHBODS a 48% record against the TAB and a 51% record against Centrebet.

MoSSBODS' landed 1 of 2 overs bets with the TAB last weekend, which lifted its overall strike rate on overs bets to 30%, still well behind its 63% strike rate on unders bets.

The bottom-line of all those numbers is that MoSSBODS, while doing no better than a coin-toss if we look at every game, does considerably better when it lands on the unders side and the overlay is greater than 6 points.

On the mean absolute error (MAE) metric, Centrebet registered the lowest MAE for game margins and away team scores this week, the TAB registered the lowest MAE for game totals, and MoSSBODS registered the lowest MAE for home team scores. 

Centrebet still leads, however, on the season-long view for home team scores and game totals, and has drawn dead-level with the TAB on game margins. The TAB narrowly wrested best MAE away from MoSHBODS on away team scores.