Bitten By the Dogs

Well, 22 tall and athletic Dogs wearing Red, White and Blue guernseys and pushing for a top 4 spot dug the hole on Saturday, and 1 short and stocky white dog wearing nothing but a green collar just about filled it in on Sunday evening - with a little help mostly from Hope.

At least, that was the story for most Investors this weekend. Investors with the Recommended Portfolio wound up losing 1.4% and other Investors lost amounts ranging from 0.4% to 9.3%. MIN #002 was the only Investor to record a profit, jumping 7% courtesy of the Hope Funds' impeccable 2 from 2 record.

Other Funds were less successful than Hope, though Chi-squared did also eke out a profit, rising just over 4c on 2 wins from 4 bets. New Heritage dropped over 9c also on 2 wins from 4 bets, Prudence dropped 5c on 2 wins from 5 bets, and Line Redux dropped 1c on 2 wins from 4 bets. All told the Funds won just 10 bets from 19, so dropping a percent or two is probably not such a bad result, especially when you consider that 3 of those 'winning' bets were on the drawn Essendon v Lions game.

What generally hurt the majority of Funds this weekend was the unlikelihood of many of the results. Only 3 favourites won and another only drew, leading to the season's most unpredictable set of results, as measured by surprisals.

The table below lists the average outcome surprisals per game for each of the rounds this season alongside which I've shown the returns recorded for each of the Funds.

At the foot of the table I've calculated the correlation between the average outcome surprisals and the Fund returns for each round. The negative correlations on the left show that the New Heritage and Prudence Funds have prospered when results have been most predictable and surprisals have therefore been low, and the positive correlations on the right show that the Hope, Chi-squared and Line Redux Funds have prospered when results have been least predictable and surprisals have therefore been high.

The latest round was a perfect example of this phenomenon. The average outcome surprisals per game was a season-high 1.42 bits, the Hope and Chi-squared Funds made profits, the Line Redux Fund almost broke even, and the New Heritage and Prudence Funds made losses.

(For information on team-by-team outcome and margin surprisals visit MAFL Stats.)

Unpredictability also affected the MAFL tipsters' predictive accuracy. The round's best performance was 4.5 from 8 as was recorded by Shadow, CTL and STM I. The worst was just 1.5 from 8 and was recorded by STM II.

Silhouette now leads our tipping competition on 108 from 152 (71%), ahead of Shadow, CTL and BKB, all on 104 from 152 (68%).

Shadow, Silhouette and STM II remain strongly in profit on level-stake wagering on their head-to-head tips, whether you start counting from Round 1 or from Round 6.

Though ELO managed just 3.5 from 8 on head-to-head tipping, it bagged a perfect 8 from 8 on line betting, moving it to 85 from 152 for the season (56%), a record that would have yielded +8.22 units from level-stake wagering. Chi correctly predicted only 4 from 8 on line betting.

Exceptional line betting generally results from exceptional margin prediction and this is indeed how you'd describe ELO's margin prediction performance this season. Its Mean Absolute Prediction Error is now just 28 points per game and its Median Absolute Prediction Error is 22.5 points per game.

In comparison, Chi's Mean APE is 29.7 points per game and his Median APE is 26 points per game. More relevantly, BKB, always the benchmark, has a 27.8 points per game Mean APE and a 23.5 points per game Median APE.

So, ELO is just 0.2 points per game behind BKB on Mean APE and is a full 1 point per game ahead of BKB on Median APE.