2013 : Round 12 - Results

Wagering can be such a cruel pastime. When Collingwood led by almost 9 goals at the main break in Sunday afternoon's game against the Dogs, I thought that our line bet on them, giving 39.5 points start, was close to bankable. I'd mentally written-off the two SuperMargin wagers, and a quick calculation revealed that the loss of these two SuperMargin bets, coupled with the successful line bet, would leave the Recommended Portfolio at virtual breakeven for the round. Not a great result, but not too bad.

When the Pies led by about the same margin at three-quarter time, I'd no real reason to change my opinion.

But the Pies then contrived to lose the final term by 5 goals to 2, eventually winning by only 34 points and, in so doing, flipping the line bet from a winner to a loser and, as a very small consolation, flipping one of the SuperMargin wagers in the other direction.

That was the one bright spot about the weekend's wagering: the success of the Bookie_9 "insurance" component of the Margin Fund. Bookie_9's 3 from 4 performance was enough to cushion what would otherwise have been about a 9c loss for the Portfolio.

In the end the loss was just on 5c, leaving the Recommended Portfolio now down by slightly less than 18c.

TIPSTERS AND PREDICTORS

All six favourites won this weekend, making for a complete absence of movement at the top end of the Leaderboard for the Head-to-Head Tipsters. Combo_NN2 still leads outright on 78.5 from 102, two tips ahead of seven other Tipsters on 76.5. Home Sweet Home still props up the bottom of the ladder, clinging desperately to a better-than-chance score.

There's also been no change in the order atop the Leaderboard for the Margin Predictors, which now sees RSMP_Weighted leading RSMP_Simple by about 40.5 points (which is equivalent to 0.39 points per game across the 102 games played so far this season). Bookie_3 lurks in 3rd, just another 3.5 points further adrift. Best performed this week amongst the Margin Predictors were the two H2H_Unadj Predictors that each turned in mean absolute prediction errors (MAPEs) of around 15.3 points per game.

It was yet another good round for all Margin Predictors, however, with none registering a MAPE above 21.78 points per game (Combo_NN2) and only three registering MAPEs above 19 points per game.

The margin prediction record of Bookie_3 serves as an interesting study at the moment. It's only predicted a margin within 6 points of the correct one in 9% of games, making it the Predictor with the poorest record on this metric, but it's also only predicted a margin more than 42 points from the correct one in 19% of games, making it the Predictor with the best record on this metric. Nonetheless, it's been 'nearest the pin' on 14 occasions, the second-most of any Predictor, and 'furthest from the pin' on 12 occasions, the third-most of any Predictor. All of which has added up to a 3rd placing on the Leaderboard.

RSMP_Weighted and RSMP_Simple continue to be the best Predictors to follow for line wagering purposes. Each has a 58% winning record in this form of wagering.

The Head-to-Head Probability Predictors all turned in stunningly good probability score performances for the round but these performances were all about equally good leaving their rankings largely unchanged bar the relegation of WinPred back into last place behind the two H2H Predictors.

In sharp contrast, the Line Fund algorithm recorded its worst probability score for a single round this season, an especially poor result when you consider that the round was one-third shorter than every other round this season excepting last week. Small wonder then that the Line Fund is down by over 26c this season.

(By the way, for those of you who've been searching for MARS, Colley, Massey and ODM Rating updates, I'll be publishing these after next week's round at which point all teams will have played the same number of games during the season.)

SUPERMARGIN PERFORMANCES

Bookie_9 has, as I already mentioned, redeemed itself this week, and now has a SuperMargin record more similar to its peers'. It has now selected the correct bucket in 13 games this season, 8 of them when predicting a home team victory, which are the ones that Investors most care about. By way of comparison, Combo_NN2 has selected the correct bucket in 14 games overall and in 8 when it has predicted a home team win.