Matter of Stats

View Original

2013 : Round 23 - Results

You want to face the Finals with a win under your belt - unless, apparently, you're the players and/or coaching staff of the Pies, the Swans, Port or, in particular, Freo - so it's nice that the Recommended Portfolio finished the season with a small profit this week, breaking a run of losses that had extended across three rounds.

Still I find myself miffed that the Dogs, having led by a margin at half-time sufficient to cover the spread on their Investor-supported Line bet, spent a second half displaying the footballing equivalent of their mascot species' sniffing of behinds, peeing on lamp-posts and rolling in things of indeterminate, malodorous but - to a canine - alluring aroma, kicking 3.5 to their opponent's 7.4 and, as a result, consigning Investors' line betting wager to the loss column.

So it is that I type this week advising of a 0.4c gain for the Recommended Portfolio rather than a gain of over 4c. That small gain leaves the Portfolio up by 7.4c on the season.

The Head-to-Head and Margin Funds both went winless for the round, dropping 0.3c and 8c respectively in the process, while the Line Fund's two correct wagers from three - the Dogs wager being the frustrating third - produced a 2.1c gain for the Fund. As a consequence, the Head-to-Head Fund is now up by just over 10c on the season, the Margin Fund is up by just under 10c, and the Line Fund up by almost 6c.

During the Finals the wagering of all three Funds will reduce further still on the basis that, historically, MAFL funds wagering in the Head-to-Head, Line and SuperMargin markets have done less well in this part of the season.

TIPSTERS AND PREDICTORS

The decision by WinPred and its family of related Head-to-Head Tipsters to opt for a Saints victory over Freo was enough to propel them into a joint share of victory in that portion of the MAFL Leaderboard, giving them an 8 from 9 performance and drawing them level with Combo_NN2 to finish the home-and-away portion of the season on 147.5 correct tips from 198, or 74.5%.

Seven other Tipsters - comprising the ProPred and (most of) the Head-to-Head families - all finished in equal 5th just a single tip further back, while two other Tipsters, Bookie_9 and H2H_Adj_7, separated the cream of this year's MAFL Head-to-Head tipping elite from Bookie Knows Best, relegating BKB to equal 14th, a position as lowly as I can recall for this most credentialled of Tipsters. This year, in fact, so run-of-the-mill was the head-to-head tipping of the TAB Bookmaker that the simple expedient of tipping the team that was higher on the competition ladder at the time the game took place - CTL's strategy - performed equally as well. 

The Heuristic Tipsters, collectively and relatively, had a quieter year, though five of them nonetheless would have returned a profit on the year to anyone who had level-staked their predictions whenever they foreshadowed a Home team victory: Shadow (+1.9 units), Easily Impressed I (+2.5 units), Easily Impressed II (+11.3 units), Short-Term Memory I (+3.5 units), and Short-Term Memory II (+3.1 units).

At the table's other end, Home Sweet Home had a particularly unsuccessful season, landing just 109.5 (55%) correct predictions. That's better than chance, but not by a wide margin.

Moving next to margin prediction, RSMP_Weighted, in its inaugural year, finished comfortably 1st amongst the Margin Predictors, ahead of RSMP_Simple, which was also in its rookie year. They were the only two Predictors to record a season-long mean absolute prediction error (MAPE) below 27 points per game. None finished with an MAPE worse than 30 points per game.

Combo_7 did enough this week to finish in 3rd place on the MAFL Leaderboard ahead of Bookie_LPSO by a little over 2 points across the 198 games of the season. Just 3 behinds or a single goal in a single game that benefitted Bookie_LPSO at the expense of Combo_7 would have been enough to reverse that finishing order.

It's interesting to note that the Margin Predictors finishing in first and in last positions on the MAFL Leaderboard were each nearest to the actual final margin of all Predictors on 8 occasions. Also, Combo_7 managed to finish 3rd despite being nearest the correct margin only twice; being furthest away only once obviously helped a great deal.

Five Margin Predictors - those filling the first four places on the MAPE ladder plus Combo_NN2 - finished the home-and-away season with line betting records sufficient to generate a positive ROI had all their predictions been wagered on at $1.90.

Bookie_LPSO took out the title amongst the Head-to-Head Probability Predictors though it had to withstand a superior final-round performance by Bookie_RE to do so. The best probability score for the round was attributable to WinPred, however, though it wasn't good enough to elevate WinPred above its 4th position.

The Line Fund algorithm finished with its second-best performance of the season in a single round to finish the year with a mildly negative probability score.

SUPERMARGIN PERFORMANCE OF MARGIN PREDICTORS

No Margin Predictor could manage more than a single correct bucket-selection in SuperMargin wagering this week. In fact, only one could manager fewer than that, Bookie_9 being the only Predictor not to select the correct bucket in at least one contest.

That leaves ProPred_3 as the Predictor with the best season-long SuperMargin performance, having selected the correct bucket in 28 games, once more often than Win_7 managed, and twice more often than Combo_NN2. These three Predictors, along with Win_3, were the only Predictors to finish the home-and-away season with a positive ROI in the SuperMargin market.

Focussing solely on the performance of Predictors when they foresaw a Home team win or draw we find that ProPed_3, Win_3, Win_7 and Combo_NN2 were the only predictors to have recorded a positive ROI - in Combo_NN2's case a profoundly positive ROI.