2012 Round 2 Results : A Little Surprisingly Unsurprising

But for the Cats' remarkable victory on Monday afternoon, this round would have been remarkable most of all for its predictability. Even with the mildly surprising Cats win, the average surprisals per game, at just 0.436 bits, was the second-lowest we've seen in a home-and-away round in the past 12 months. That's what you get when 8 of 9 favourites win, 7 of them priced at $1.33 or less.


Since most of the MAFL tipsters had wisely piled onto the favourites, this made for a great round of tipping, with all but a handful of Head-to-Head Tipsters bagging 8 from 9, and the average tipster registering 7.9 from 9.

That leaves 13 Head-to-Head Tipsters tied for the lead on the MAFL Leaderboard, all on 14 from 18, and all a tip ahead of the TAB Bookmaker. Even the worst Head-to-Head Tipster, Home Sweet Home, has correctly picked over 60% of the results so far.

Margin prediction proved more challenging than winner prediction, due almost entirely to the fact that four games ended with victory margins in excess of 90 points. Only Bookie_3 produced a sub-30 MAPE round, finishing with an MAPE of 29.87 points per game. Next best was Bookie_9 on 32.92 points per game, with Combo_7 a tick back on 32.97 points per game. The All-Predictor average for the round was 39.1 points per game, a full 10 points worse than the average for Round 1.

These results leave Bookie_3 in the outright lead across all Margin Predictors, a little over 1 point per game ahead of Combo_7, with Bookie_9 about another half a point per game further back.

Following the Margin Predictors for the purposes of line betting has been generally unrecommended this season, the exception being Bookie_3, which has so far predicted 56% of all line betting results. Only a contrarian strategy would have proved successful in relation to the predictions of most other Margin Predictors, most notably in the cases of Combo_NN1 and H2H_Adjusted_7 where such an approach would have yielded a 72% success rate.

Probability scoring in the head-to-head market has also proven startlingly simple in the early rounds. In the most recent round, all Head-to-Head Probability Predictors registered probability scores of between 0.55 and 0.60 per game. This has left WinPred leading the pack on 0.37 per game and a little ahead of ProPred on just under 0.34 per game. At this point, the TAB Bookmaker is languishing in last place with a score of 0.31 per game - a score which is about three times the historical average. This shall not last ...

The Line Fund currently has a 9 and 9 record for predicting the ultimate line winner, but has attached probabilities to its predictions in such a way as to accumulate a slightly positive probability score.


Despite finishing very close to a result in 3 of the weeks' SuperMargin contests (which, as I noted here, is not an especially remarkable outcome), Investors finished the round with an 0 and 8 record. This leaves the Margin Fund down 20% and the Recommended (aka Only) Fund down 4%. On the assumption that we'll select the correct bucket about 14% of the time, the probability that any 8 selections will produce no winners is about 30%.

As it turns out, Investors, had they wanted to pursue the current policy of wagering only when a Predictor tipped a Home team victory or a drawn game, would have been better off wagering on the predictions of any of the H2H-based Predictors, as shown in the following table.

Here you can see that a unit wager on the predictions of HU3, HU10 or HA3 whenever they tipped Home team victories or draws, would have yielded a profit of $18.50, while a similar approach to the predictions of HA7 would have yielded a profit of $11.50. These are the only Margin Predictors for which, so far, this approach would have ground out a profit.

If, instead of wagering only on the bucket corresponding to a Margin Predictor's prediction, you also wagered on the buckets on either side (excluding the drawn game bucket) - a policy I considered for some time as the one appropriate for the Margin Fund - then a profit could have been spun from the suggestions of Bookie_9 (+$6.50), Win_3 (+$2.50) and Win_7 (+$2.50). Had we pursued such a strategy for the Margin Fund we'd have currently been down by only 5% rather than by 20%.

Perhaps you're more interested in wagering on Away teams, in which case the following table is of greater interest.

Here we see that the predictions of Bookie_3, Bookie_9, ProPred_7, Combo_7, and (ironically) Combo_NN2 are the only ones that have been profitable so far, and even then only if you'd decided to wager on the buckets either side of these Predictors predictions.

For completeness, here are the combined results for every Margin Predictor, regardless of whether it predicted a Home win, Away win or drawn game.