2012 Round 5 Results : Precision Wagering

We're in profit.

Not, as was surely more likely, due to any astute wagering by the Line Fund, but instead as a result of two remarkable SuperMargin results. On Saturday night I watched in disbelief as the Roos translated a 14 scoring shot advantage into only a 34-point win over the Suns, in so doing piloting the result into the heart of the 30-39 point range we were requiring. That was worth 6c to Investors (not, I apologise, the 8.25c I'd shown in the Ready Reckoner previously).

Sunday night then produced an even greater miracle as Adelaide deftly delivered a 19-point victory, thereby landing both the Line and the SuperMargin wagers for us, a result that necessitated a final margin somewhere on the 14-19 point isthmus. That was worth another 5.25c.

Other, smaller collects came courtesy of GWS' losing by only 42 points to the Dogs, and Melbourne losing by only 16 to the Saints.

All told, the Line Fund made just 3 successful wagers from 7, dropping 6.5%, while the Margin Fund had 2 profitable tilts at the market from the 4 it attempted, climbing 40% on the back of them. Overall Portfolios therefore grew 4.75% across the round and are now up 2.75% on the season.

CN2, the predictions of which determine Margin Fund wagers, is having another remarkable year of margin prediction. It's now selected the correct bucket in 8 of 45 games, in 4 of 24 when it's predicted a Home team win - and hence we've wagered - and in 4 of 21 when it's predicted an Away team win.

Four other MAFL Margin Predictors have also selected the right bucket in 8 games - 3 of the Head-to-Head based Predictors and Combo7. The full set of results are shown below.

There are a couple of extra rows in each block of data this week as I've included the results for one further strategy. Previously, you'll recall, I've shown the returns that would have been achieved by wagering on a Predictor's preferred bucket only, or by wagering on its preferred bucket plus the two buckets immediately adjacent. The new rows pertain to the strategy of wagering on the preferred bucket plus only one of the neighbouring buckets, that being the bucket nearest the actual margin predicted by the Predictor.

So, for example, if the Predictor tipped a team to win by 12 points then the wagers would be on the 10-19 and the 1-9 point buckets, since the 1-9 point bucket is closer to 12 points than is the 20-29 point bucket.

The new strategy is labelled "Preferred + Nearest Bucket" in the table.

What prompted me to consider this strategy was the fact that, this week, the head to head prices posted just prior to locking in our wagers caused CN2 to switch buckets in two games: the Roos v Suns and the Adelaide v Port games, as it happens. Whilst the outcome of each switch was fantastically favourable to us on this particular occasion, I was haunted pre-game (and, moreso, mid-game) by how I'd feel if the buckets that CN2 was originally predicting wound up being the correct ones. In future seasons, assuming the Margin Fund survives, one way to ameliorate this risk regret would be to wager on the preferred and the nearest bucket.

So far though, the strategy of wagering only on the preferred bucket is the superior strategy for most Predictors.

The Margin Predictors' generally strong showing in the SuperMargin market is reflected in their individual and collective MAPE performance on the MAFL Tipster Dashboard.

Every Margin Predictor produced an MAPE for the round of under 20 points per game, and the all-Predictor average was a remarkable 16.44 points per game, the lowest I can remember. Best was ProPred3 on 14.00 points per game and worst was Bookie3 on 19.51 points per game.

Despite this relatively poor showing across the season Bookie3 remains in 3rd place on 27.41 points per game, behind only Combo7 on 27.25 and Bookie9 on 26.35.

The line betting performance of the Predictors also continues to comfortably exceed chance, with every Predictor bar CN2 now sporting a season-long line betting accuracy of greater than 50%. Win3 and Win7 have the best statistics having correctly selected 64% of line betting winners so far this season. The all-Predictor average for the round was 6.8 from 9.

Head to head tipping was similarly accurate, the all-Tipster average coming in at 7.3 from 9 thanks largely to the fact that 8 of the 9 favourites were winners this week. Many Tipsters, along with BKB, scored 8 from 9, the only Tipster significantly lowering the average being Home Sweet Home, which correctly predicted only 3 from 9.

That leaves us still with a logjam atop the Head-to-Head Tipster Leaderboard with 13 Tipsters tied on 37 from 45 or around 82%.

It was also a week for accumulation amongst the Head-to-Head Probability Predictors, each of them recording positive results yet again. Best performed was WinPred, which racked up 0.51 Pieces per game (for a discussion of the Pieces unit, please see this blog). WinPred remains in 4th place however, behind the two Head-to-Head Predictors and ProPred, but ahead of the TAB Bookmaker.

The Line Fund algorithm also turned in a net positive performance on probability scoring but remains marginally negative across the whole season.