Matter of Stats

View Original

2014 - Round 27 : Swans (Just)

Almost all of MoS thinks the Swans will win on Saturday.

ChiPS, not for the first time, is the only dissenting voice, but it has no say in MoS wagering, so Investors find themselves, unusually for a Grand Final, with a full ballast of bets. A 4% Head-to-Head wager on the Swans lays beside a 0.6% Line wager on the Swans giving 7.5 points start, both nestling alongside two SuperMargin wagers that form the season's final Chasm of Death. If you didn't know better, you'd say it was scripted ... oh, wait ...

Though there's much activity this week, it again amounts to relatively little in terms of overall risk and reward due to the restrictive caps that have been in place throughout the Finals series. So, even a best-case outcome - a Swans victory by 8 or 9 points - yields less than a 2c gain, and a worst-case outcome - a Swans defeat - threatens less than a 2c loss.

TIPS AND PREDICTIONS

ChiPS, contrarian to the end and unlike MARS, Rates the Hawks more highly than the Swans, and so grants Hawthorn the Home Ground Advantage. The three point advantage the Hawks accrue as a result adds to their five point Rating advantage and their six point Interstate Status filip, leaving the Hawks as 13-point favourites according to ChiPS once a final, small deduction for the Hawks' inferior Form going into the GF is included.

That leaves C_Marg conspicuously out on its own amongst the Head-to-Head Tipsters this week - and not even Home Sweet Home would have been enticed to join C_Marg had it still been active this late in the season.

Given the otherwise widespread conformity of opinion, however, we're now able to declare Bookie_9 as the Best MoS Head-to-Head Tipster of 2014, and Combo_7, Win_7, Win_3 and WinPred as joint Runners Up.

Bookie_9 will finish the season at the head of the Leaderboard with a 72% accuracy rate. Curiously, if Sydney lose, Bookie_9 will record exactly the same score this year that it did last season when it finished five tips adrift of the winner, Combo_NN_2. Combo_NN_2 sits seven tips behind Bookie_9 this season. Twelve months is a long time in (tipping) football ...

C_Marg will finish either level with or two tips ahead of Combo_NN_2, which is not too bad for a Tipster with no Bookmaker input whatsoever. It would, obviously, be all the more impressive had Combo_NN_2 performed as well this year as it did last. 

C_Marg stands out too amongst the Margin Predictors, its forecast of a 13 point Hawks victory 18 points or more different from every other Predictor. Win_7 sits at the other extreme, tipping the Swans to win by 28 points, a couple of points more than any other Predictor and about 16 points more than the average.

The resulting range of margin forecasts defined by C_Marg's and Win_7's predictions - 41 points - is quite wide, and this diversity of opinion extends beyond just these two Predictors as reflected in the 10 point standard deviation of the predictions. As a group, the predictions are, however, virtually skewless, the mean and the median predictions equalling +12.3 and +12.4 points respectively. 

Combo_7's and Bookie_LPSO's margin forecasts differ by less than a point, so the latter will not be able to catch the former on the MoS Leaderboard. Accordingly, we can also declare Combo_7 as the Best MoS Margin Predictor of 2014. Its likely final Mean Absolute Prediction Error (MAPE) of around 28.4 points per game is a little higher than its 26.9 points per game performance of 2013 when it finished fourth behind the two RSMP Predictors and Bookie_LPSO. I think it's fair to say that game margins have been, objectively, harder to predict this season than last. Bookie_LPSO will finish second this year, and Bookie_9 third.

C_Prob, like C_Marg amongst the Head-to-Head Tipsters and Margin Predictors, stands out amongst the Head-to-Head Probability Predictors, its 35% assessment of the Swans' chances over 20% points lower than any other Predictor. WinPred rates the Swans' chances highest at 85%, which gives us a whopping 50% point range and contributes to the 17% point standard deviation in the collectively probability assessments for the game.

The three directly Bookmaker-derived Probability Predictors have probability assessments that are sufficiently similar that no possible result on Saturday can change their final ordering on the MoS Leaderboard, which will therefore see Bookie-OE take out the MoS Best Probability Predictor title, Bookie-LPSO finish second, and Bookie-RE finish third.

These same three Predictors filled the podium positions at the end of last season too, though then the ordering was Bookie-LPSO, Bookie-RE and Bookie-OE. As well, average Log Probability Scores were much higher last season than they have been this season because, relatively speaking, the results this year have been more surprising. That measurably true in an information-theoretic sense since the average Surprisals per game in 2013 was 0.759 bits per game while this year it's been 0.836 bits per game.

The Line Fund algorithm has finished the year with a very confident prediction, rating the Swans as 70% chances to cover the 7.5 points start they're giving the Hawks. Generally, the algorithm has produced sharper predictions this year compared to last, the average probability assessment of the algorithm's favoured team this season coming in at 61.2% compared to last season's 58.6%. The increase in sharpness hasn't resulted in a reduction in calibration however, since the algorithm's average Log Probability Score will finish much higher this year.