2015 - Round 23 : The Ins and Outs Of It

/

It's been anything but a usual week in the wagering markets for the AFL at the TAB, with speculation surrounding - and then actual - team changes leading to frequent and prolonged suspensions of some markets, significant delays in the posting of others, and a final set of prices fairly different from what we would have seen had teams been playing their regular squads.

The table below shows the Head-to-Head and Line markets at three points in time:

  • firstly, mid Monday, which is when it's customary for all Head-to-Head markets to be available. Even at this point the TAB Bookmaker was feeling squirrely about the Port v Freo game and wasn't willing to speculate about a price
  • next, early on Tuesday afternoon, which is when it's customary this year for the Line markets to be posted. Now the Tigers v Roos game is causing some alarm - very curiously, only for line wagering - and the Port v Freo game remains problematic
  • lastly, on Thursday evening, just after the teams had been announced. At that point there was some lingering concern about the Cats v Crows game, which was resolved just a short time later with the Head-to-Head market being posted at $2.25 / $1.65 and the Line market at Geelong +13.5 and $1.70 / $2.15

The MatterOfStats rule is that the Fund algorithms and various Tipsters and Predictors are required to use as their inputs, if they need them, the Head-to-Head and Line market prices for any game as they are at the time the Line market is posted for that game (or as near to that time as I'm able to catch them). That rule has resulted in the various Funds and Forecasters using the prices shown in bold in the table above.

Despite all the shenanigans of the week then, it's turned out that this rule hasn't much affected the inputs to the models, though MoS' Funds and Forecasters would probably rather have used the later market data for the Tigers v Roos, Lions v Dogs, and maybe the Dees v Giants, and Cats v Crows games.

Anyway, it all amounted to very little in terms of wagering, the Head-to-Head Fund passing on the round altogether, and the Line Fund making just a single bet on the Eagles giving 52.5 points start.

No need then for a Ready Reckoner then, the fate of the Overall Portfolio being easily described in a single sentence: if the Eagles win by 53 points or more, the Portfolio rises in value by 1.4c; if not, it falls by 1.5c.

HEAD-TO-HEAD TIPSTERS

Even though the Bookmaker prices ultimately changed relatively little during the week, the actual values of those prices are very different to what they would have been had teams not rested regular, generally more-talented players. These player ins and outs also mean that team Ratings, which reflect the abilities of the regular team, and team form data, which reflect the more-recent performances of these same regular players, are potentially less useful as predictors this week. 

Many of the model-based algorithms combine, in their own particular way, Bookmaker data with team Rating and venue data, and this week have found that these two types of data are sometimes telling different stories.

Which is why, for example, the WinPred, ProPred and H2H families of Tipsters have all opted for a Roos win.

Regardless, the overall level of disagreement amongst the Head-to-Head Tipsters is surprisingly low, the lowest it's been in fact since Round 18. Only three Tipsters have Disagreement Indexes above 15%: C_Marg (19%), Easily Impressed II (25%), and Home Sweet Home (41%). It's interesting to note that none of these Predictors use Bookmaker data in coming up with their selections.

Only two games have more than two Tipsters siding with the underdogs, 11 doing this in the Tigers v Roos game, and eight doing it in the Port v Freo game.

MARGIN PREDICTORS

Whilst it turns out that, the Tigers v Roos game aside, the differences between Bookmaker and team Rating data don't have much of an effect on the teams that the Head-to-Head Tipsters predict will win, these differences have had a much larger effect on the Margin Predictors.

So much so, in fact, that the week's all-Predictor Mean Absolute Deviation (MAD) is at a season-high 10.7 points per game per Predictor.

Some very unfamiliar Predictors find themselves most different from the norm this week, Bookie_3, with a MAD of 15.6 points per game, having the highest MAD for the first time this season, and a slew of other Predictors recording double-digit MADs for the first time as well.

The largest MAD has come in the Richmond v Kangaroos game where it's over 25 points per Predictor and where the predictions span a range from Combo_NN1's 62-point Richmond win to Win_7's 8-point Kangaroos victory. In the Hawks v Blues game the MAD is also greater than 4 goals and in the Port Adelaide v Fremantle game it's over 2 goals.

C_Marg, which currently leads ENS_Greedy by almost 122 points for the season, has a margin prediction 33 points different from ENS_Greedy's in the Tigers v Roos game, and a margin prediction 35 points different in the Port v Freo game. Unfavourable results in those games could see C_Marg's lead dramatically reduced.

PROBABILITY PREDICTORS

Input misalignment has also had a significant effect on the Head-to-Head Probability Predictors resulting in an all-Predictor MAD of 6.0% points per Predictor per game, which is the highest such MAD since Round 16, and the 6th-highest MAD of the season.

As we saw for the Margin Predictors, the effects are largest in the Tigers v Roos (23% point MAD), Port v Freo (8%), and Hawks v Blues (7%) matchups.

We find here too that the list of high-MAD Predictors includes some unfamiliar names, most notably Bookie-LPSO and Bookie-RE.

C_Prob, which currently heads the MoS Leaderboard, has most to lose if Richmond, Adelaide or Port Adelaide win, all of which are Bookmaker favourites. This week is likely to be decisive in C_Prob's quest to outpredict the TAB Bookmakers across the entire season.

The Line Fund algorithm, perhaps also befuddled by the confusing nature of its inputs this week - though, with its season history, that's a difficult call to make - rates five teams as better than 60% chances on the line market this week, all of them playing Away: the Kangaroos (74%), Carlton (73%), Essendon (67%), the Western Bulldogs (61%), and Fremantle (60%).

FORECASTER DISAGREEMENT

The unusual pattern of Margin and Probability Predictor disagreement this week is evident in the table below: