Matter of Stats

View Original

2017 - Round 15 : The Impossible 9

If every contest in a 9-game round was essentially a coin toss, the chance of correctly tipping all nine winners would be 511/1. This weekend, based on the bookmaker's mid-week odds, the chance is only about one-eighth of that or about 68/1. Those are the longest odds we've seen for picking the card all season, ahead of the 55/1 for Round 1 and the 40/1 for Round 4. (The shortest for any of the 9-game rounds by the way was 18/1 for Round 3).

That said, even in the games that are priced close to parity in the betting markets, the Head-to-Head Tipsters have all still tended to agree, with the result that the round's Overall Tipster Disagreement Index is just 22%, which is just 2% above the all-season average.

There are, though, three credible dissenting voices in the Suns v Roos game, where the two ladder-leading RSMP Tipsters and C_Marg (well, okay, two credible dissenting voices) have chosen the underdog Roos.

More remarkably, there are seven dissenting voices in the Dockers v Saints game where, as at the time of writing, the Dockers enjoy narrow favouritism. In that game, BKB is joined only by Home Sweet Home in tipping the Dockers. 

Amongst the Margin Predictors, opinions are extremely aligned. The all-Predictor Disagreement Index is only 3.1 points per Predictor per game, which is yet another season low. 

And. for the first time since Round 5, C_Marg finds itself with only the second-largest mean absolute deviation (MAD), this week trailing MoSHBODS_Marg and roughly equal with MoSSBODS_Marg.

From a game-by-game view, only the round's final contest between Fremantle and St Kilda has elicited an all-Predictor MAD of more than 4 points. In that game the range is defined by MoSSBODS_Marg's prediction of a Saints win by 16 points and Bookie_9's of a Dockers win by 4 points. In no game does the range of margin forecasts span more than 20 points.

Before we move on the Head-to-Head Probability Predictors, I thought it'd be interesting to have a quick look at which teams have been hardest to create accurate margin forecasts for this season.

The table below records each Margin Predictor's mean absolute error (MAE) for every team when playing as the designated home team, away team, or either. The columns are ordered left to right based on the Predictors' all-team MAE, and the rows in each table are order based on the all-Predictor average MAE for each team within the block.

One thing that the tables make very clear are the main sources of C_Marg's relative inaccuracy this season: Hawthorn, the Kangaroos, Geelong and St Kilda at home, and the Kangaroos, Melbourne, West Coast and Hawthorn when playing away.

It's also interesting to see the range of overall MAEs across the 18 teams, with the all-Predictor average for Richmond just under 22 points and that for Adelaide almost exactly twice that size.

Finally, note that the MoS twins' superiority has come mostly from being more accurate with the harder-to-predict teams rather than more accurate with the easier-to-predict teams.

Anyway, returning to the Head-to-Head Probability Predictors the story this week is also one of broad agreement, though a little less so than last weekend. The all-Predictor MAD this week is 3.2% points per Predictor per game, the second-lowest value for a round for the season behind last week's 2.6% points.

MoSSBODS_Prob has the round's highest MAD of 4.3% points per game, ahead of MoSHBODS_Prob (4.1%), and C_Prob (4%).

Three games have a MAD above 4%, the highest coming in the Dockers v Saints game where it is 6.3% per Predictor. That game also has the largest range of probability estimates for Fremantle (from 34% to 53%), though five other contests also have estimates spanning at least a 12% point range.

WAGERS

There's a lot of activity this week, especially in the head-to-head markets.

The Head-to-Head Fund has six wagers totalling just under 11% of original funds, while the Line Fund has two wagers totalling just under 5%. Five of the six head-to-head wagers are priced at $2 or under, so the upside in most games is fairly modest.

The majority of that upside belongs to St Kilda, who could lift the Overall Portfolio by almost 2% should it overcome Fremantle on Sunday. They also hold the bulk of the weekend's downside, however, and could shave about the same amount off the Overall Portfolio price with a loss by 3 points or more.

In total, just over 4% of the Overall Portfolio is at risk, and a best-case set of results is capable of adding about 3.6% to that Portfolio.

MoSSBODS AND MoSHBODS

Here are the MoS twins' views on team and total scoring for the week.

Commentary to follow.