2018 - Round 17 : Marking Time

Two games this weekend have an expected margin in single digits, and two more have expected margins of around two goals. The largest expected margin - for the Dees v Dogs game - is about 33 points, which leaves the majority of games in the two to four-and-a-half goal range and has produced the lowest average TAB Bookmaker margin since Round 13. It's 17.9 points per game this week.

That's a couple of points per game under the 2018 season average and about 3.5 points per game under the average for Round 17s since 2012.

Still, as we've noted in passing previously, actual average margins can generally be expected to be higher than expected average margins, a phenomenon we can quantify by recognising that the actual margin for a game roughly follows a Normal distribution with a mean equal to the expected margin and with a standard deviation somewhere around the mid- to late-30s.

Based on that assumption, the chart at left maps the relationship between the expected margin for a game and its average actual margin for a range of expected margins.

It shows, for example, that for games where the expected pre-game margin was 20 points, the average actual margin is likely to be in the 32 to 34 point range, depending on the assumption we make about the relevant standard deviation.

We can see that, across all Expected Pre-Game Margins, the Average Actual Margin exceeds the Expected Pre-Game Margin for all Expected Pre-Game Margins shown here, although the gap shrinks towards zero as the Expected Pre-Game Margin increases. At an Expected Pre-Game Margin of 50 points, for example, the Average Actual Margin is about 52 to 54 points.

Put simply, if you legitimately think that the best estimate of the pre-game margin is a Team A win by X, the average actual margin is likely to be greater than X, especially if the expected pre-game margin is small.

You can get an intuition for why this might be the case by imagining that you think the home team is a legitimate 5-point favourite in a particular game. In that case, there are only 11 margins that will be less than or equal to your predicted margin - anything from a 5 point away team win to your predicted 5 point home team win - but there are a huge number of larger wins that, combined, account for a lot more of the probability. These will drag the overall average above 5 points.


It's another largely dissension-free weekend for the Head-to-Head Tipsters, with only the MoS twins opting underdog in the Roos v Swans game, C_Marg doing the same in the Crows v Cats game, and Home Sweet Home doing the same, mostly ineffectual thing in both of those games and in three more.

Most importantly, there's complete agreement amongst those at the top of the Leaderboard, so we'll again see no movement there.

There are also generally low levels of disagreement amongst the Margin Predictors, with only MoSSBODS_Marg (4.6) and C_Marg (4.1) producing mean absolute disagreement (MAD) figures above 4 points per game.

Looking across the matches, only the Gold Coast v Essendon game has produced a MAD of over 4 points per Predictor. The MAD there is 4.4, elevated especially by C_Marg's relatively low margin forecasts for the away team win.

In comparison to the current leader, Bookie_Hcap, a best set of results for MoSSBODS_Marg would see it move from about 61 points behind to about 16 points behind, while a best set of results for MoSHBODS_Marg would see it move from about 127 to about 83 points behind. Most likely, as the blog post title suggests, this will be a round for marking time ...

The mean expected margin across the nine games and all Margin Predictors is 17.3 points per game, a bit more than half a point a game lower than the TAB bookmaker's average,.

Turning lastly to the Head-to-Head Probability Predictors, we find the largest MAD in the probability estimates for the Gold Coast v Essendon,  GWS v Richmond, and Adelaide v Geelong games, where they range from 4.0% to 4.4% points per Predictor. The largest range of estimates comes in the GWS v Richmond game, where it is 21% points, spanning a MoSHBODS_Prob low of 25% to a MoSSBODS_Prob high of 46%. It's no fun when family members disagree.

MoSSBODS_Prob, which now holds top spot on the Leaderboard, again has the round's largest MAD, this week at 4.2% points per game, and has the most extreme estimate in five of the contests. MoSHBODS_Prob has the next-largest MAD of 3.9% points per game and the most extreme estimate in four contests.


It's another restrained set of recommended wagers this week, though that didn't help Investors finish last weekend with a profit. We've four wagers from MoSHBODS in the head-to-head markets this week, the two largest over 3% on the Cats and the Tigers, and just two from MoSSBODS in the line markets, the largest only 1.2% of the Fund.

(Please click on the image below to access a larger version.)

Combining head-to-head and line wagers where appropriate, the largest swing is in the Crows v Cats game where a worst-case result would strip 1.1c from the price of the Combined Portfolio, and a best-case result would add about the same amount.

No other game carries a swing of more than 2%, although the situation in the Giants v Tigers game is especially interesting because MoSSBODS and MoSHBODS have wagered on different teams. There, the best case for Investors is a Tigers win by 1 to 14 points, which would simultaneously secure MoSHBODS' head-to-head bet on Richmond, and also MoSSBODS' line bet on GWS with 14.5 points start.

In total, 3.1% of the original Combined Portfolio is at risk and the maximum upside is just over 3%. 

To finish, here are MoSHBODS' and MoSSBODS' opinions about likely team scoring.

Commentary about these forecasts will come in the week's Overs/Unders blog.