It's been a while since we've reviewed the history of game margins and, in today's blog, we'll consider that history from a number of perspectives.
Firstly, here's the more traditional points-based view, with the top lines (in red) tracking the season-by-season average absolute final margin, the more jagged of the lines providing the raw data and the dotted line tracking a 9-season, centred, moving average.
Both the raw data and the moving average data hint at an increasing trend in final margins in recent seasons, with the raw 2015 result reversing the short-term declines of 2013 and 2014 relative to the two seasons prior. Even with those declines, the five most-recent seasons have all managed to produce average absolute margins above 35 points per game, while the season before that, 2010, produced an average of 34.995 points per game.
The last time that the competition went a whole season with a sub-30 average victory margin was 1976, so unless 2016 is a remarkable aberration from trend we look set to make it 40 straight seasons since that mark was achieved.
It's possible, though unlikely, that the higher margins in recent decades have presented a misleading picture of changes in the underlying abilities of the teams. Maybe there's been no change in the spread of team abilities, and final margins have just happened to increase by chance. To investigate this possibility I've also charted the average pre-game margin expectations provided by MoSSBODS 2.0 Ratings (adjusted for venue effects).
For reasons explained in this earlier post on blowout results, even unbiased expected absolute margins will tend to be lower than actual margins. We see that in the chart above, but we also see that the expected absolute margins follow a similar trajectory to actual absolute margins (the correlation is +0.75). The 2015 uptick in actual margins was not, however, matched by a similar increase in expected margins, so it might be the case that this year's higher average margin was not a consequence of a widening in the average difference in team abilities.
In any case, the solid green line tells us that the average difference in team abilities (after adjusting for venue effects) has been about 20 points or more since 2008, and that three of the five greatest differences from the past 30 seasons came in 2011, 2012 and 2013.
Scoring is, of course, influenced by Conversion Rates, which we know have varied across history (see this blog for a chart of them). To control for this we might analyse game margins in terms of Scoring Shots rather than Points, which is what's been done for the chart below.
The trajectories shown here are consistent with those from the earlier chart, and the correlations are +0.79 for the two Actual time series and +0.94 for the two Expected time series.
It's interesting to note that, since about the mid 1960s, the average Actual Scoring Shot difference has mostly tracked in a relatively narrow range from about 8 to 10 Scoring Shots and that, during this same period, the Expected Scoring Shot difference has mostly tracked in the 4 to 6 Scoring Shot range.
So, to conclude and summarise, both the Scoring Shot and the earlier Points perspectives suggest that:
- The era of declining margins that started shortly after the "modern" form of the draft was introduced in 1986 appears to have ended about 10 or so seasons ago
- Since then the trend has been generally upwards, though only slightly, and perhaps a little distorted by the introduction of Gold Coast and GWS
- The most recent season, 2015, whilst registering higher Actual Margins than 2013 and 2014 was, according to MoSSBODS, associated with lower Expected Margins. Regardless, the trend, even for Expected Margins, is generally upwards since about 1995.
The next few seasons are going to be critical in determining whether victory margins have plateaued or whether they'll continue their gradual increase.