An Adelaide upset win was this week the key to a bountiful haul for Investors and, once that result went unachieved on Friday night, albeit only narrowly, taking over 6.5c of the Recommended Portfolio with it, the remainder of the round was then solely about sizing the eventual loss. At least, that's how I saw it.

That loss wound up being only about 2.5c and could easily have been a most unexpected profit had the Roos not finished so poorly against the Saints, leaving them just half a point short of covering the spread. A lazy behind for the Roos late in the game would have added 3.4c to the final result for Investors and left us up by about 1c on the round. On such apparent insignificances are seasons won and lost.

Still, it's hard to complain too much about that Roos result when we benefited from a late Lions goal that turned a 3.1c loss into a 1.6c gain. So, as I've often written before, swings and roundabouts ...

Across the round, the Head-to-Head Fund dropped over 14c, most of it on the Crows, while the Line Fund gained 3c and the Margin Fund 1c, its lone winning wager coming courtesy of the Roos' lethargy in the final term, which marooned their winning margin at the upper end of the 50 to 59 point range.

The Portfolio loss of 2.5c for the round leaves it now up by 7.4c on the season. So, ultimately, this week was just a flesh wound. I reckon it was worth it for the thrill of the Adelaide wager. YMMV.

Incidentally, while I was out on a long walk following the Dees' thumping on Saturday, I was convinced that the Line Fund had done worse this year on teams receiving start than on teams giving start. I was even pondering some suitable adjustments to the Fund for next year. As yet another example of the fallibility of human memory - well of mine at least - and the power of confirmation bias, it turns out that I was completely wrong. This season we're 20 and 12 (63%) on teams receiving start, and 23 and 19 (55%) on teams giving start. In a fight between recorded evidence and human memory, back the former every time.

TIPS AND PREDICTIONS

Three of the lesser Heuristic Tipsters fared best on Head-to-Head Tipping this week, Short Term Memory II, Easily Impressed II and Shadow all bagging 8 from 9. Tipsters nearer the top of the ladder managed only 6s and 7s though, and the all-Tipster average was exactly 6.

Bookie_9 still leads the group, two tips ahead of Combo_7 with Bookie_3 and now Combo_NN_1 one tip further back. Combo_NN_1 was the only Tipster to correctly predict the Dons' win over the Pies.

Best amongst the Margin Predictors was Bookie_3, whose 23.55 Mean Absolute Prediction Error (MAPE) was slightly superior to Bookie_LPSO's 23.69 and RSMP_Weighted's 23.90. These were the only Predictors to record MAPEs of under 4 goals. The all-Predictor average MAPE was 28.36 points per game, slightly lower than the all-Predictor average MAPE across the season, which now stands at 29.41 points per game.

Combo_7 heads the Margin Predictor race, now just over 10 points clear of a threatening Bookie_LPSO, and almost 54 points clear of Win_3. C_Marg remains last, though its performance this week has given it renewed hope for overhauling Combo_NN_2 and finishing only second-last.

Nine Margin Predictors have profitable season-long Line betting records, including C_Marg, who has now chosen the right side of the coin on 53% of occasions. The all-Predictor average Line betting record stands at 53.9%.

Speaking of profitable records, three Margin Predictors have season-long profitable records on SuperMargin wagering - Win_3, Win_7 and H2H_Adj_3 - though the third of that group is only barely in the black.

Amongst the Head-to-Head Probability Predictors, C_Prob recorded the round's best probability score, though all Predictors returned highly positive results. C_Prob currently lies fourth on the ladder, just a fraction behind Bookie-LPSO and also not far behind Bookie-RE. Bookie-OE still heads the ladder.

The Line Fund algorithm recorded another negative probability score for the week, though this week's score was less negative than last week's. Nonetheless, the algorithm's season-long average probability score is now negative. That's the price you pay for being sharp without (always) being right.

Looking instead at the Margin Predictors' performances only when they've predicted a home team win or draw, eight of them are currently in profit and a ninth, H2H_Unadj_10, is exactly at breakeven. Unfortunately, none of the nine Predictors includes Bookie_9 or Combo_NN_2, which power the Margin Fund.

This week's best SuperMargin performances belonged to Bookie_LPSO, the two RSMP Predictors, Bookie_3, and Combo_NN_1, all of which selected the correct bucket twice.