The TAB Bookmaker is predicting a round of moderate to comfortable wins for seven of its favourites this week, all seven expected to win by between about two-and-a-half and six goals according to their line market handicaps. The two exceptions are the Roos who are expected to win by only about a goal-and-a-half on Friday night, and Collingwood who are expected to win by slightly less than a goal on Sunday.

Reinforcing this tendency to firmer favouritehood is the fact that, across the nine games, the average line handicap is 24.2 points per game, the second-highest average for a round this season, behind only the 28.3 points per game average for Round 3.

Five of the week's underdogs are home teams, three of them priced attractively enough, according to the Head-to-Head Fund, to warrant wagers. The biggest of these wagers is on the shortest-priced of the value underdogs, a 3.5% bet on the Tigers at $2.20 to beat the Pies, and the smallest is on the longest-priced, a 1.4% bet on the Dogs at $4.35 to upset Fremantle.

The three head-to-head bets total 7.1% of the Head-to-Head Fund, which is only slightly less than the proportion of the Line Fund exposed by its three wagers, two of them on teams for which Investors also have Head-to-Head Funds exposure (the Dons and the Dogs), and the third the only line bet on a favourite, one on the Swans giving 21.5 points start to the Cats.

This week's Ready Reckoner then sees risk concentrated, naturally enough, in the Dons and the Dogs, who between them could be responsible for almost 6c of profit or about 4.5c of loss.

Sydney and Richmond are the only other teams who can affect Investor fortunes, both capable of altering the value of the Overall Portfolio up or down by about 1.5c. Combined, the week's maximum possible gain is 8.8c and maximum possible loss is 7.3c.

HEAD-TO-HEAD TIPSTERS

Given the earlier discussion about the strength of bookmaker favouritism in this round, we might expect to see high levels of agreement week amongst the Head-to-Head Tipsters. And, indeed, we do. Not since Round 1, in fact, have the Head-to-Head Tipsters been so much in agreement, their overall Disagreement Index this week coming in at just 16%. Four of the usual contrarian tipsters have remained stoically ignorant of the perceived superiority of the favourites in a number of the contests and recorded the round's highest Index values: Home Sweet Home (43%), Easily Impressed I (30%), Short-Term Memory II (28%), and Follow The Streak (26%). 

The more-often-than-not member of the contrarian clique, C_Marg, has this week crossed over to the consensus and is in the very unusual position for it of being amongst the 10 Tipsters recording the lowest Index values of 8%.

Nine of those 10 conforming Tipsters come from the top nine places on the MatterOfStats Head-to-Head Tipster Leaderboard, which means we'll see no reshuffling amongst them this weekend. The highest-ranked Tipsters with any hope of climbing higher are from the H2H and Win families.

Despite the overall low levels of disagreement across all nine contests, the Dons v Roos game has still managed to generate some dissent, the final count finishing in favour of the Roos by 18 to 12. There also 10 dissenting Tipsters in the Tigers v Pies game, and six in the Lions v Power contest. The Hawks and the Eagles, however, are unanimous favourites, and Fremantle has been prevented from earning this same status only by the uber-contrarian Home Sweet Home.

MARGIN PREDICTORS

As I think I've noted before, the relative strength of the favourites in any given week has no necessary influence on the level of agreement amongst Margin Predictor margin predictions, since they're trying to foretell not just who will win but by how much. So, while we find that disagreement levels are low amongst the Head-to-Head Predictors this week, we find that disagreement amongst the Margin Predictors as measured by average MAD is at about the same level as last week and, at 6.7 points per game, at its third-highest level this season.

Combo_NN2 and C_Marg are, as they were last week, two of the larger contributors to that average, C_Marg distinguishing itself by forecasting the largest Hawks, Eagles and Pies wins of all the Predictors, and Combo_NN2 forecasting the largest Crows and Swans wins.

Bookie_9 has the week's lowest MAD, and MoS' Official Margin Predictor, Combo_7, has the week's second-lowest.

Combo_NN1, which currently heads up the MatterOfStats MAE Leaderboard, has a MAD of 6.1 points per game, only slightly below the all-Predictor average, while H2H_Adjusted_3, which lies second on that Leaderboard, has a MAD of 8.8 points per game. The largest difference between these two Predictors is in Friday night's game where Combo_NN1 has the Roos winning by 23 points, and H2H_Adjusted_3 has them losing by 3 points. Since the gap between the two Predictors on the Leaderboard is only 12 points, that game alone, should the result be a Roos win by less than 4 points (or a draw or a loss), could be enough to see the two swap places.

Looking at the predictions, instead, on a game-by-game basis, we find that it's the Hawks v Dees game that has the largest range of forecasts and, comfortably, the largest MAD of 13.1 points per Predictor. The two other games with above-average MADs are the Crows v Saints (8.9 points per Predictor), and the Eagles v Suns (7.6).

In all three of the games with above-average MADs the Margin Predictors are unanimous about which team they think will win; the debate in these games is solely about the margin of victory. In the Dons v Roos and the Tigers v Pies games, however, though the MADs are smaller there are six Predictors forecasting an upset victor.

The game with the smallest MAD is Lions v Power, where the average is just 3.6 points per Predictor. This is another game where the Predictors are unanimous about which team is most likely to win and, were it not for Combo_NN2's forecast of a 13 point win by Port Adelaide, every margin prediction in that game would fall within a 15 point range spanning a 23 to a 38-point Port Adelaide victory.

PROBABILITY PREDICTORS

The Lions v Power game has also generated very low levels of disagreement amongst the Head-to-Head Probability Predictors. The probability assessments for this game span only a 9% point range, which is the smallest of all nine contests, and they have a MAD of just 2.5% points per Predictor, which is second-smallest, narrowly behind the MAD in the Eagles v Suns game.

Due almost entirely to C_Prob's assessment of the home team's chances, the largest range (39% points) and largest MAD (9.5% points per Predictor) comes in the Tigers v Pies game.

Overall though, disagreement levels are very low this week - so low, in fact, that the all-Predictor all-game MAD of 5.5% per game is a season low.

C_Prob has the week's largest MAD (8.5%), almost 2% points higher than those of the two H2H Predictors, which this week offer the same probability assessments as one another, there being no need to invoke the 25% cap. Bookie_OE, which heads the MatterOfStats LPS Leaderboard and which is MoS' Official Head-to-Head Probability Predictor, has the week's low MAD of 3.4% points per game.

In the line market, the Line Fund algorithm rates three teams as 60% or better chances: Melbourne (67%), St Kilda (62%), and Essendon (60%). Three other teams, Sydney, Gold Coast and Port Adelaide, are rated at least 55% chances.

DISAGREEMENT AND SUCCESS

With all the talk of Disagreement Indexes in the previous sections, you might wonder about the relationship between Tipster and Predictor levels of disagreement and the accuracy of their forecasts.

A few quick calculations using each Tipster's and Predictor's week-by-week disagreement metrics and year-to-date performance reveal that the correlation between:

• the Head-to-Head Tipsters' average Disagreement Index and the number of games they've tipped correctly is -0.73
• the Margin Predictors' average MAD and their MAE is +0.18
• the Head-to-Head Probability Predictors' average MAD and their Log Probability Score is -0.49

The strength of these correlations, and the Tipsters or Predictors that are most outlying, is made clear when we plot the underlying data. 

None of the Head-to-Head Tipsters stands out as having been much less- or more-rewarded for contrarian behaviour than any other Tipster in that a straight line optimally drawn through the cloud of points in the chart will not find itself far away from any single point.

Amongst the Margin Predictors, however, some points would be quite distant from such a line. Win 3 and Win 7 and, to a lesser degree Combo_NN1, are three Predictors most obviously in that category. Excluding those three Predictors lifts the linear correlation to +0.36. Dropping the four H2H Predictors as well boosts it to +0.79.

For the Probability Predictors, C_Prob is the one that's most bucked the trend for greater divergence to be penalised with lower log probability scores by recording a much higher average log probability score than its average level of disagreement would imply. Excluding it from the calculation lifts the correlation between average log probability score and average MAD to -0.83.

So, to this point in the season, conformity with the consensus has, on-balance, benefitted Tipsters and Predictors, most widely amongst the head-to-head forecasters where there are no significant exceptions, and amongst probability predictors where there's only one significant exception. Amongst margin predictors the returns to conformity have been more mixed, with seven predictors finding themselves well away from the general MAE / MAD trade-off line.