2015 - Round 14 : Backing Up After A Bye

So far this season 12 teams have played after a bye in the previous week and their collective record is just 4 wins and 8 losses. I'm hoping that the two MoS Funds, who both determined to nominate last weekend as their impromptu "bye", fare a little better this weekend.

Neither Fund, as you (and certainly I) might have hoped, have remained inactive again this week, the Head-to-Head Fund venturing two bets totalling just over 5% of that Fund, and the Line Fund venturing four wagers totalling 20% of the other. In aggregate, the pair have put 14% of the original Overall Portfolio at risk, easily the largest proportion imperilled in a single round since the start of the season.

The two Head-to-Head bets mix conservatism and adventure, the first a wager of just over 2% on the $1.50 favourite Swans, and the second a 25% larger wager on the $3.75 outsider Pies.

Similarly, the four Line bets are split 2-2 in terms of start given and start offered, the two on the underdogs enjoying three- to four-goal starts, and the two on the favourites offering roughly two goal or smaller handicaps.

The Pies then represent both the largest upside and the largest downside for Investors this week, a win by them of any size promising a 5.7c increase in the value of the Overall Portfolio, and a loss by 25 points or more threatening a 4.1c decrease.

Three other teams, the Suns, Dogs and Dons, carry identical, but smaller, upsides and downsides, with favourable line betting outcomes capable of adding 2.7c to the Overall Portfolio, and unfavourable outcomes capable of stripping 3c from it. The only other team with an ability to alter the value of the Overall Portfolio is the Swans, whose loss would see the Overall Portfolio fall by just under 1c and whose victory would see it rise by 0.5c.

That leaves four wagerless contests that can be enjoyed by Investors in the same way as they probably enjoyed football before they had any financial stake in the outcomes.


This week the Head-to-Head Tipsters are as unified as they've been all season, their collective Disagreement Index (DI) of 12% the smallest it's been all year. (Note that an internal audit has revealed some small errors in the previous DI calculations for the Head-to-Head Tipsters. This anomaly has been corrected this week and the responsible party has been counselled. I'd have let him go, but this blog isn't going to write itself.)

Twenty two of the Tipsters have each recorded the season's lowest DI of 7%, an Index made possible only by the extraordinary levels of unanimity and near-unanimity about the likely outcome in most of the contests.

One game sees complete unanimity (and yes, I'm aware of the absolute nature of that concept), and in four other games there's only a single dissenter, for three of which it's only Home Sweet Home that's pointing in the other direction. In two more games, only a pair of Tipsters form the minority and in yet another there's only a trio of contrarians. That leaves just one contest, the Dons v Saints game, where there's anything approaching a concerted opposition, that game seeing six Tipsters plumping for an upset Saints victory.

With such high levels of agreement it's of no surprise at all to reveal that the top 11 Tipsters on the MoS Leaderboard have identical tips and that we need to look to Combo_NN1, the 12th Tipster on that Leaderboard, before we find a contrarian opinion. It likes Geelong over Adelaide and will, accordingly, move up to join the Win Family in equal 9th should that tip prove accurate.


The Margin Predictors have also turned in a relatively small Mean Absolute Deviation (MAD) this week, their 5.0 points per game per Predictor result the 2nd-lowest of the season.

But for C_Marg, who has reverted to type this week by turning in a 12.6 MAD - almost 4 points per game higher than the next highest Predictor - that average would also have been a season low.

C_Marg has, in particular, gone out on a limb about the victory margins of the Swans, Dogs, Eagles, Dons and Crows, for each of which teams its opinions are most extreme. No other Predictor finds itself as Predictor Most Extreme in more that three contests, the dubious honour of being almost as contrarian as C_Marg falling to Combo_NN2, which is bravest in its opinions about the Hawks' and Tigers' victory margins, and mildest in its opinions about the Roos'.

Just two games have a MAD above a goal, the Freo v Lions game at the high end with the only double-digit MAD, and the Tigers v Giants game the other in that pair and with a MAD of 7.7 points per Predictor. The Crows v Cats game has the round's lowest MAD of just 2.7 points per Predictor.


The Head-to-Head Probability Predictors are also finding it difficult to disagree this week, their collective MAD of 3.5% points per game per Predictor a season low, despite being dragged higher by the freshly-confident C_Prob whose prognostications are, on average, 7% points different from its average peer. 

As for C_Marg, most aberrant are C_Prob's assessments of the Swans, Dogs, Eagles, Dons and Crows, for each of which teams it finds itself more enthusiastic about their respective chances than does its colleagues.

In contrast, the two H2H Predictors have the round's smallest MADs.

Looking on a game-by-game basis we find that it's the Swans v Power contest generating the highest levels of dissent, albeit that the MAD for this game is only 5.5% points per Predictor. The Dockers v Lions game has engendered the lowest levels of dissent, the MAD for this game coming in at least 1.1% points per Predictor and the range spanning only 4% points from a near-certain 93% to a nearer-still-certain 97%.

The three most directly Bookmaker-based Predictors have registered very high levels of agreement this week, which means that only C_Prob might realistically alter the landscape at the top of the Mos Leaderboard.

Meantime the Line Fund algorithm has assessed only one team, Essendon, as having a 60% shot at line betting success this weekend, its next-highest assessments of 59% and 57% being reserved for the Lions and Dogs respectively.


Below is the corrected table of Disagreement data for the Head-to-Head Tipsters and the updated data for this and all other Tipster and Predictor groups.

The linear correlations between Disagreement and performance now stand at:

  • -0.53 for Head-to-Head Tipster Accuracy and DI
  • +0.134 for Margin Predictor MAE and MAD
  • -0.359 for Head-to-Head Probability log probability score and MAD