In the first four games of Round 12 underdogs came close to springing upsets, finishing within 2 goals of at least a draw in all four contests. Two of these valiant underdogs were responsible for two of the three successful wagers Investors made this weekend.

On Friday, St Kilda, in kicking a goal after the siren, pulled within 4 points of the Crows, landing the SuperMargin wager we had on that game, then, on Saturday, the Giants stayed close enough to their opponents to comfortably secure the Line bet we had on them with over 10 goals start. The Saints gallantry, however, also saw off Investors' Line bet on the Crows giving 16.5 points start.

The Hawks produced the week's only other successful wager by defeating the Lions by more than the 48.5 points start they were asked to give. In the end the Hawks covered the spread by a little too much - their late goal, which extended their lead to 65 points, being enough to deny our SuperMargin bet on them to win by 50-59 points.

Other unsuccessful wagers were a line bet on the Eagles who, despite generating 29 scoring shots to their opponent's 19 contrived to win by only 10 points, and two other SuperMargin wagers.

Overall the gains outweighed the losses and Portfolios grew by 1.25c on the round to now be down for the season by just under 8c.

The underdogs' ultimate failure on Thursday, Friday and Saturday, coupled with Sunday's losses by the two remaining underdogs, allowed favourites to sweep the round, which was of course great news for our Head-to-Head Tipsters given their generally conservative leanings this season. On average they predicted 5 from 6 correctly this week with the worst performances being turned in by Easily Impressed I and II, each managing just 3. Eleven Tipsters remain co-leaders, all now on 79 from 102 (77.5%), and the TAB Bookmaker still sits two tips from the lead.

Margin prediction was also riotously straightforward, so much so that the Margin Predictors produced their 3rd-lowest all-Predictor MAPE for the season at just 19.04 points per game. Amongst them, ProPred_7 was the most astonishingly accurate, on average predicting within 12.5 points of the eventual margin and prevented from returning an even more astonishing result only as a consequence of predicting the Tigers to win by 33 points.

Worst amongst the Predictors was Bookie_3, which registered a MAPE of 26.97 points per game. Despite this it retained 3rd spot on the season-long ladder on 28.87 points per game, headed still by Combo_7, now on 28.42 points per game, and Bookie_9, now on 28.16 points per game.

Accurate margin prediction once again led to successful line market predictions too, the all-Predictor average for the round coming in at 4 from 6 and leaving all Predictors with better than 50% season-long records.

It also led to accurate SuperMargin bucket selection for the week as most Predictors found themselves within a bucket or two of the correct bucket in at least half of the week's contests.

Six Predictors - including Combo_NN2, I note disapprovingly - remain profitable if we assume that we'd wagered on their selected bucket in every contest. Win_3 and Win_7 continue to produce near-misses at an extraordinary rate. Win_3 has predicted within a bucket of the correct result in 38 of the season's 102 games, but in 31 of those games it's predicted the bucket adjacent to the correct one. Win_7's results are only marginally less exasperating. It's been within a bucket in 37 games but in the bucket adjacent to the right one in 29 of them.

The Head-to-Head Probability Predictors also racked up impressive results for the round, all generating large and positive probability scores. ProPred did best, but the differences in probability scores amongst all Predictors were very small, so their was little movement on the Leaderboard where the TAB Bookmaker still leads fairly comfortably.

Once again, the Line Fund algorithm registered a mildly negative probability score for the round. We know from previous statistical analyses, however, that line betting using an algorithm producing negative probability scores can still be profitable even when Kelly-staking is employed. To investigate this I calculated the ROI that would have accrued from wagering on the Line Fund algorithm's line market predictions in every game this season where its level of confidence justified a wager on one team or the other. For the season overall so far its ROI is -12.6%, which comes from an ROI of -3.6% when wagering on Home teams and -37.3% when wagering on Away teams.