As we venture into the world of score prediction for the second-last time this season, we find near-perfect harmony amongst our four forecasters.

At this time of year it’s reasonable to expect close contests, but in the West Coast v Melbourne Preliminary Final the Eagles’ venue advantage is only narrowly offset by the Dees’ underlying superiority. At least that’s what the MoS twins think.

Richmond’s lead over Melbourne rose a little (to 0.6 Scoring Shots) on MoSSBODS, but fell a little (to 1.3 points) on MoSSBODS, as the two Systems reacted slightly differently to Melbourne’s 104 points to 71, and 24 scoring shots to 21, victory over Hawthorn.

It was another good week for most forecasters of all forms, with all but two of the Head-to-Head Tipsters selecting both winning teams, the average mean absolute error (MAE) for the Margin Predictors coming in at 11.3 points per game, and large, positive probability scores for all of the Head-to-Head Probability Predictors.

Last week’s four finals produced only 545 points amongst them, or just 68 points per team per game, with the highest-scoring game offering only 22 goals and 27 behinds for a total of 159 points.

As I write this, the bookmakers have Melbourne and Collingwood as favourites for the upcoming semi-finals, but only by about 13 and 9 points respectively, so there’d be no great surprise were there to be upsets in both games.

Richmond preserved their narrow leads on both the MoSSBODS and MoSHBODS Team Rating Systems, remaining just 0.3 Scoring Shots ahead of Melbourne on MoSSBODS, and dropping 0.2 points on MoSHBODS where their lead is now only 1.5 Points.

Historically, especially in recent years, Finals are where the MoS models tend to do best, and this week saw the MoS twins tip all four winners, a feat that no other Head-to-Head Tipster managed.

If scoring was low during the regular, home and away season, it's probably unreasonable to expect it to ramp up dramatically, if at all, during the Finals, and none of our forecasters are predicting any such profligacy. They're suggesting that 80 to 82 points per team is a reasonable expectation for this week.

It seems very early, I know, to be posting forecasts about Finals that are over a week away, but one of the advantages of player-ignorant models is that they are, well, ignorant about players.

Richmond finish the home and away season atop both MoSSBODS and MoSHBODS, though with the smallest leads they've had for some time - just 0.3 Scoring Shots on MoSSBODS, and 1.7 Points on MoSHBODS - after a round in which they narrowly prevailed over the far-lower ranked Western Bulldogs, dropping ratings as a consequence. Melbourne, in contrast, registered a solid win over the similarly-rated GWS both in terms of points and scoring shots, which served to lift their rating on both Systems.

It was another 7 or 8 tip, sub-30 mean absolute error (MAE), positive probability score weekend for most forecasters, the end of which saw relatively little movement on the respective Leaderboards.

There's quite a range of expected totals across the last nine games of the home and away season. For the MoS twins they cover 154 to 183 points - about a 5 goal range - and for the bookmakers they cover 137.5 to 180.5, which is a 7 goal range.

Round 23 is very much a round of two halves, with four of the contests expected to be decided by about two-and-a-half goals or less, and the other five by between 4 and 13 goals.

In the same way that accuracy can be a poor estimator of a forecaster's ability, raw wins, draws and losses can be a poor estimator of a team's ability, so it always amuses at this time of year how readily some commentators will conflate where a team sits of the ladder with how well they've played across the season. There really are such things as lucky wins and unfortunate losses, and sometimes the less able team wins and the better able one loses.

Only Consult The Ladder and Home Sweet Home tipped fewer than seven winners this week, which left ENS_Linear still with a two tip lead over Bookie Knows Best at the top of the Leaderboard.

ENS_Linear is now on 134.5 from 189 (71%).

If, as it seems, the mean total score for an AFL game in 2018 is 165 to 170 points, last weekend was a textbook example of regression to the mean after the relative scorefest that was Round 20.

Just three games this week are expected to finish with a margin under 3 goals, but seven with a margin under 5 goals, which has made for an all-game average of about the same as last weekend's 24.5 points per game. This week, it's 24.2 points per game.

There was very little movement on either MoS System this week, with MoSSBODS swapping Sydney into 10th and dropping Brisbane Lions into 12th, and MoSHBODS moving Essendon ahead of Geelong, and the Western Bulldogs ahead of the Brisbane Lions. For the remaining teams, there were rating changes, but they were relatively small.

The MoS twins were very bullish about the Dees' chances this weekend and paid the price in all three of our forecaster fields.