Matter of Stats

View Original

2015 - Round 25 : Following The Script

This week sees the two Home teams comfortable favourites on the TAB at the time of writing, especially the Hawks, whose $1.30 price tag makes them unavailable for wager for the Head-to-Head Fund. It's responded by placing a moderately sized wager on the other Home team, the Swans, who at $1.55 the Fund sees as exceptional value.

The size of that wager, 3.6%, is the largest single head-to-head wager on a team since Round 17, and is all the more surprising given the tighter cap restrictions that the Fund faces in this part of the season. What's driving its magnitude is the H2H algorithm's assessment that the Swans are 77% chances.

The Line Fund, unconstrained in its choice of team but completely constrained in its bet size, has opted for a lone bet on the Hawks, giving 22.5 points start, representing 0.6% of the Fund.

Two bets makes for a very straightforward Ready Reckoner, which shows that a Hawks line win would add 0.3c to the Overall Portfolio, while a loss would subtract 0.4c from it, and that a Swans head-to-head win would add 0.8c, while a loss would snip 1.4c. 

Victories by Hawthorn and Sydney would preserve the perfect records that teams since 2000 finishing 3rd have had when playing teams finishing 7th, and that teams finishing 4th have had when playing teams finishing 8th in Semi-Finals.

Losing Qualifying Finalists generally, in fact, have a fairly imposing record over that same period, having won 26 of the 30 contests that have taken place. Three of those four losses have come from teams finishing 3rd, and none from teams finishing 1st or 2nd.

Winning Elimination Finalists, of course, in aggregate have a mirror 4 and 26 record. Amongst that group, though, it's teams from 6th that have by far the best record having secured three of the only four wins created.

Teams from 6th also have the best record In Week 1 of the Finals amongst teams finishing 5th through 8th. They've now won, including last week, 11 of their 16 contests against the team finishing 7th. In the other Elimination Final, which pits 5th versus 8th, the record is surprisingly split 8 and 8.

With 1st's and 2nd's dominant records over 4th's and 3rd's respectively, that makes the 5th versus 8th contest the only one of the four in which the lower-finishing team has won at anything like a 50% rate.

HEAD-TO-HEAD TIPS

The Head-to-Head Tipsters, as if scholars of this history, have unanimously selected the Hawks and the Swans to win this week.

That means, of course, that there'll be no changes on the Leaderboard and that, with just three more games to be played after this week, C_Marg can be declared MoS Head-to-Head Tipster of the Season.

C_Marg can also, if it picks correctly in the five remaining games, finish on 150 correct tips, which is one-and-a-half tips better than last year's winner, Bookie_9, which achieved its score of 148.5 from one additional game. (Tipsters were not credited with any score in this season's Adelaide v Geelong game.)

It can't, however, eclipse Combo_NN2's 153.5 score from 2013, nor H2H_Unadj_10's mammoth 162.5 score from 2012. In hindsight, those were incredibly predictable seasons - in the literal sense of the word.

MARGIN PREDICTIONS

There's unanimity too amongst the Margin Predictors about the more likely victor in each game, but there is conjecture about the size of the victories, especially Hawthorn's.

In that game, the all-Predictor Mean Absolute Deviation (MAD) has come in at 5.7 points per Predictor, and the Range spans over 5 goals, topped by Combo_NN2's 53-point prediction, and tailed by the 21-point predictions of Combo_7, Bookie_9, H2H_Unadj_3, and H2H_Unadj_10. 

The other game sees a MAD of just 3.8 points per Predictor and a Range of just over 2 goals.

C_Marg, whose lead over ENS_Greedy now stands at about 110 points could, worst case, see that gap whittled to about 90 points, while ENS_Greedy's 9-point lead over RSMP_Weighted could be shrunk to just 2 points if both results prove unfavourable for it.

Whilst C_Marg's lead in the competition won't be seriously imperilled this week, its sub-30 Mean Absolute Error (MAE) could be. To retain it, C_Marg's aggregate absolute error (AAE) over the two games needs to be below 64 points. No other Margin Predictor is capable of finishing this round with a sub-30 MAE, though ENS_Greedy could finish the season with such a MAE in the unlikely event that its AAE over the five remaining contests comes in under 46 points.

RSMP_Weighted could also achieve the feat with an AAE below 37 points, as could RSMP_Simple with an AAE below 13 points. 

PROBABILITY PREDICTIONS

C_Prob, which leads all-comers in the Head-to-Head Probability Predictor race, will surrender its lead over two of the Bookie trio this weekend should the Crows upset the Hawks. It rates the Hawks as 87% chances while the Bookie family have them as between 73% and 76% chances, and a loss by the Hawks will see C_Prob slip to 3rd behind Bookie-OE and Bookie-RE.

That Hawks v Crows contest has the smaller of the week's MADs and Ranges, the MAD in the Swans v Roos game much larger at 8.0% points per Predictor (vs 4.9% points), and the Range a little larger at 18% points (vs 14% points).

WinPred and ProPred define the upper ends of the Ranges in the two contests, while Bookie_OE, in an unaccustomed role, defines the lower end of the Ranges for both games.

In the line market, the Line Fund algorithm has been fairly conservative this week, assessing the Hawks at 54% chances, and the Roos as 51% chances. Whatever the results then, the Line Fund algorithm won't be making much of a dent in its season-long sub-zero log probability score.