The off-season always seems a good time for adopting a more sweeping historical perspective in the analyses here on MatterOfStats. Today we're going to be reviewing Scoring Shot Conversion rates across the 119 seasons of the VFL/AFL from both a venue and a team perspective.

For clarity, a reminder that a the Conversion Rate is defined for current purposes as Goals / (Goals + Behinds), where Behinds include Rushed Behinds. Whilst it might be desirable to exclude this latter form of Behind-scoring from the analysis since their prevalence doesn't entirely reflect goal-kicking accuracy, the data necessary to do so doesn't exist in the historical record.

Let's start by reviewing the all-team, all-venue Conversion Rate time series (see chart at right) where we find a long-term growth trend interrupted by a decline from about 1935 to about 1960 and by another decline over the past 15 years or so. Across the history of the sport, seasonal Rates have ranged from a low of 38.1% in 1900 to a high of 55.2%, fittingly exactly 100 years later, in 2000.

During this period, games have been played at 44 different venues, the idiosyncrasies of which when coupled with the goal-kicking accuracy levels of the era, have combined to produce patterns of Conversion Rates unique to each venue. In the heatmap below we colour-code based the average Conversion Rate for all games at a particular ground in a given season.

(Note that 11 of the 44 venues do not appear in this heatmap because in no single season were at least 3 games played at that venue.)

The tendency for Conversion Rates to, broadly, grow over time is reflected in this chart by the movement from bright green to dark red as we move left-to-right across time. Grounds such as Princes Park, Victoria Park, and the MCG provide especially interesting case studies of the combinatory effects of the specific abilities of the teams playing most often on them during a season and the general levels of accuracy prevailing across all teams and venues at the time.

Other grounds such as the Gabba where games have been played within a single era of similar goal-kicking accuracy better isolate the unique influence of the one or two teams playing most often on the ground during the period. York Park is another example of such a ground, though caution needs to be exercised in interpreting its data, as for Stadium Australia and the WACA, because as few as 4 or 5 games were played at these venues in some seasons.

Because of the relatively large range that Conversion Rates have spanned across history and the relatively narrow range they've traversed in more-recent times, variability across venues for the modern era is difficult to discern in this heatmap. To accentuate these differences a little more then, let's constrain our attention to the 1980 to 2015 period.

The York Park anomaly, though certainly discernible in the previous heatmap, is now especially noticeable, the period 2010 to 2015 clearly different from the pre-2010 era. Docklands also catches the eye as a ground on which, it appears, Conversion Rates tend to be above-average.

Even with a narrowing of our historical perspective to only the last 36 seasons, it remains difficult to determine which of the venues have yielded consistently above- or below-average Conversion Rates, so for this purpose I've created another heatmap, which subtracts from each venue's Conversion Rate for a given season the all-ground average for that season. If, for example, the Conversion Rate in 1962 at a particular ground was 49.6%, while the all-ground average for 1962 was 50.5%, the value we'd use for the following heatmap would be -0.9%.

In this heatmap, the lighter the green colour the lower the relative Conversion Rate and the darker the red colour the higher the Conversion Rate. White conveys a Conversion Rate consistent with the all-ground average.

Now we can see, for example, that:

• Since about the mid-1950s, the MCG's previous record of producing above-average Conversion Rates has been replaced by a record of producing average or below-average Conversion Rates
• Subiaco has, recently, passed through three distinct phases of increasing relative Conversion Rates
• Docklands has consistently produced average or above-average Conversion Rates
• Football Park, the Gabba and Carrara have tended to witness below-average Conversion Rates

It would be an interesting exercise for a football historian to piece together the history of some of these grounds - how their shape, configuration, surface and surrounds differ and have changed - and to attempt to relate this history to the data.

The Conversion Rate history of a particular ground will, of course, be most influenced by the team or teams that call that ground their home, so it might also be instructive to review this history on a team-by-team basis, which we do in this next and final series of heatmaps.

Here too in this all-time team-by-team chart, as was the case for the venue-by-venue view, the overall look is heavily influenced by the broad trends in Conversion Rate history, though a handful of teams do stand out for particular seasons, for example the Dons of 1918 and 1921, and the Hawks of 2014 and 2015.

Again, as a way of highlighting the most-recent history we'll next confine ourselves to the 1980 to 2015 period. This brings the performances of particular teams in particular seasons into much sharper relief.

We can see, for example, the stretch of relatively high Conversion Rates churned out by the Hawks from 2000 to 2006 (during, it must be said, a period of generally elevated Rates), and by the Dogs from 2005 to 2011 (when Rates were less universally raised). Also evident is the decline in the Dees' Conversion Rates over the past four seasons.

Finally, let's review the season-normed version of the team-by-team all-time data, again showing above-average performances in red, below-average performances in green, and about-average performances in white.

Now we see interesting and strong season-to-season correlations in many team's relative Conversion Rates. The Dogs, for example, with only a few exceptions, have tended to convert at a rate higher than the all-team average in every season since about 1997, possibly aided by the relatively large number of games they've played at Docklands (195). Sydney have also had an impressive conversion record over this period, though less-impressive than the Dogs' and apparently having fallen off in the past couple of seasons.

Richmond, in contrast, managed only two seasons of above-average Conversion in the 27 season period from 1987 to 2013. Mercifully for their supporters, they've strung together back-to-back above-average seasons for the first time since 1986.

Hawthorn's record is particularly amazing, they having recorded above-average Conversion Rates in 28 of the last 34 seasons, and in 17 of the last 20. Essendon also had an exceptional run of above-average seasons, their record being 23 from 29 such seasons across the period 1980 to 2008 (and including nine above-average seasons in succession from 1982 to 1990). Since then, however, the Dons have struggled, producing only 2 above-average seasons in 7 attempts.

Across the entire history of the sport, below-average performances have tended to persist moreso than above-average performances. A team that recorded a below-average Conversion Rate in one season has had about a 59% chance of repeating such a performance in the season following, while a team that recorded an above-average Conversion Rate in one season has had about a 54% chance of repeating it the next year. In interpreting these figures note that, because of the typical skewness in Conversion Rates across teams within a season, about 53.5% of teams across history have recorded above-average Conversion Rates while only 46.5% have recorded below-average Conversion Rates.

This team-by-team view of Conversion Rate history also presents an interesting opportunity for students of the game to look at the characteristics of different teams in different eras - their composition, playing style, coaching staff and home ground - in an attempt to identify those features which might have been contributory to extended periods of kicking accuracy or inaccuracy.