Average Points per Team per Game in Portions of the H&A Season

In this chart we look at rates of scoring in the early, middle, and late parts of the home-and-away seasons of recent years to investigate the possibility that scores tend to be higher or lower in particular portions.

Average scores are higher in the first 5 rounds of the home and away season than in the middle (as defined here) in 34 of the 45 seasons (76%). They are also higher in the last 5 rounds of the home and away season than in the middle (as defined here) in 28 of the 45 seasons (62%).

Effect of Days Rest (and Travel) on Performance Relative to Expectation

It’s often contended that teams playing on fewer days’ rest are at a disadvantage, and the following chart seeks to test that hypothesis by looking at how teams have performed relative to how the MoSHBODS Rating System (which does not make adjustments for days rest) expected.

For this chart, Excess Performance is defined as a team’s actual margin in a game (ie its score less that of its opponent) less its expected margin according to MoSHBODS.

For most teams, having less than 7 days between games appears to have little to no effect on their performance relative to expectation. North Melbourne and Sydney are the obvious exceptions. As counterexamples, Gold Coast , GWS and Western Bulldogs appear to do relatively better on less rest, although Gold Coast and GWS have relatively small sample sizes

Note that some teams have, on average, non-trivially over-performed (eg Geelong, Hawthorn and Sydney) or under-performed (eg Gold Coast and West Coast) relative to MoSHBODS expectations, so we should focus on the ordering of the dots (and the error bars) for a team rather than the absolute values.

Follow-up Analysis - Differential Rest and Differential Travel

A reasonable response to the analysis above would be that, firstly, a team’s opponent’s rest matters too, as does the amount of travel that both the team and its opponent need to undertake in order to get to the venue.

To address both of these issues we’ll fit the following linear regression:

lm(Excess Performance ~ Differential Days Rest:Team + Differential Travel:Team)

We’ll again define Excess Performance using MoSHBODS, but this time we’ll exclude Venue Performance Values from the calculation because these implicitly carry a distance component to the extent that distance does indeed affect teams’ performances at any given venue.

The table at right summarises the fitted model.

From it we can deduce that:

(Looking firstly at the left-hand block of numbers)

  • Having more rest than an opponent is not a statistically significant advantage for any team (and, in any case, the effect sizes are very small - the largest estimate advantage is 0.6 points per extra day for Sydney)

  • The only statistically significant Days Rest related coefficients are negative, which implies that the team tends to do better with fewer days rest than its opponents, This is the case for Gold Coast, Western Bulldogs, West Coast, Carlton, and Richmond (which is broadly consistent with the analysis above). As an example - and make of it what you will - but Gold Coast is about 2.4 points better off for every day of rest fewer it has compared to its opponent.

(Looking next at the right-hand block of numbers)

  • Every team’s excess performance is reduced to a statistically significant extent by the amount of travel it had to undertake to arrive at a venue compared to its opponents.

  • Port Adelaide’s performance is most affected by travel. It declines by 0.9 points for every additional 100km of travel.

  • Conversely, Richmond’s performance is least affected by travel. It declines by only 0.3 points for every additional 100km of travel.

Finally, to provide some context for those travel-related numbers, the table at left records the average distance travelled per home-and-away game for each team.

Broadly speaking then, we have:

  • Western Australian teams c 1,300km

  • Queensland teams c 750-900km

  • NSW teams c 500-600km

  • South Australian teams c 500-550km

  • Victorian teams c 300-450km

Winning and Losing Rates in Close and Non-Close Games

Another common assertion is that “good teams know how to win the close ones” so in the chart that follows we investigate teams’ win rates in close and non-close games over an extended period.

For the purposes of this analysis, we define a close game as one that was won by 12 points or less. Every other game we’ll define as “not close”. We’ll include all games in the analysis, including Finals.

We find, firstly, that most teams win the close games at about a 50% rate, the exceptions being:

  • Under 50%: Adelaide, Gold Coast (on a small sample)

  • Over 50%: Geelong

We also find that the majority of teams have a closer to 50% record in close games than they have in non-close games

  • True for: Brisbane Lions, Carlton, Collingwood, GWS, Geelong, Gold Coast, Melbourne, North Melbourne, Port Adelaide, St Kilda, Sydney

  • Opposite is true for: Adelaide, Richmond, West Coast, and Western Bulldogs

  • Distance to 50% is about the same in Close and in Non-Close games: Essendon, Fremantle, and Hawthorn

Change in Team Conversion Rates After Wins and Losses in the Home-and-Away Season

How do teams' conversion rates (Goals / Scoring Shots) tend to respond after a win and after a loss? Since teams that win tend to convert at slightly better-than-average rates, and teams that lose tend to convert at slightly worse-than-average rates, one hypothesis would be that teams' conversion rates will regress towards the mean, falling after a win and increasing after a loss.

The initial hypothesis is very clearly borne out: for every team, across the 25-season period, wins tend to produce less accurate kicking in the next game, while wins tend to produce more accurate kicking in the next game.

Relationship Between Scoring and Winning Rates

Scoring Shots

Here we’re looking at the relationship between the number of scoring shots that a team generates and its winning rate, and how this relationship varies by era.

We can see that the relationship is S-shaped for all eras, but steeper in some eras than in others.

Also, fewer Scoring Shots are associated with a 50% win rate (about 24) in the 2000-2024 era than in the 1980-1999 era (about 28 or 29)

Finally, increasing from 25 Scoring Shots to 29 Scoring Shots (ie just one Scoring Shot per quarter) in the 2000-2024 era is associated with about a 22% point increase in Winning Rate (from 57% to 79%).

Goals

Here we do the same thing, but use goals instead of scoring shots.

We see that the relationships are also S-shaped for all eras and, again, steeper in some eras than in others.

Also, fewer Goals are associated with a 50% win rate (about 13) in the 2000-2024 era than in the 1980-1999 era (a little over 14).

Finally, increasing from 14 Goals to 16 Goals (ie just one Goal per half) in the 2000-2024 era is associated with about a 20% point increase in Winning Rate (from 61% to 82%)

Relationship Between Total Score and Victory Margin by Era

It seems reasonable to expect that games with larger victory margins might tend to also produce larger total scores, but it would certainly be possible to argue for other relationships between these two metrics.

We know that scoring generally has varied quite marked across history, so we’ll conduct this analysis by splitting history into six eras.

We find that there is, indeed, a positive relationship between game margins and total scores in each of the six era.

Further, in the modern era, higher Margins have been associated with slightly lower Total Scores than in the previous era. For example, the average Total Score for a Margin of 20 points has been 173.2 points in the modern era, whereas it was 191.7 points in the previous era.

Lastly, having been fairly steady for about 80 years at an 0.43 to 0.47 points increase in total score for every 1 point increase in margin, in the modern era that multiplier has dropped to 0.34. Put another way, it’s now more common to have larger margins with smaller total scores than has been the case in previous era.

Relationship Between Performance in Consecutive Home-and-Away Games Relative to Expectation

The term “momentum” is tossed around relentlessly in sporting commentary (and never with a well-define meaning, in my opinion), and one way it might manifest is from one game to the next where an above-average performance relative to expectations one week might tend to be associated with another above-average performance relative to expectations one week (and ditto for a below-average performance).

To test for this phenomenon we’ll create a scatter of successive home-and-away performances relative to MoSHBODS’ expectations for every team for games between 2000 and 2024.

No team shows any substantial evidence of a relationship between its performance relative to expectations in one game and its performance relative to expectations in the previous game (as evidenced by the tiny R-squared values).

Another interpretation of this finding is that, in truth, teams’ underlying ratings don’t change from week-to-week, but that MoSHBODS makes adjustments to teams’ ratings (and hence expectations) that entirely account for and are necessary to explain any such “momentum" in team performances.

Close Games and Blowouts

One way that seasons are sometimes characterised by fans revolves around the number of close games, and the number of blowouts, that were seen during the season.

In this next chart we’ll look at the proportion of each type of game for each of the season, defined a “close game” as one won by less than a goal, and a “blowout” as one won by 5 goals or more.

Something that immediately stands out is the increase in the frequency of blowouts since about 1980 (although there were a few earlier seasons with similarly high proportions of such games). A contributing factor to this higher prevalence of blowouts is the generally higher scoring in recent times, especially in the 1980s and 1990s, so there might be a case for redefining what a blowout is in earlier seasons.

In contrast, the proportion of close games has generally remained in the 0 to 20% range, with figures in the early seasons very similar to those from recent times.

Looking at some recent and other extremes:

  • In 2015 only 14 of 206 (7%) games finished with a margin under a goal and 109 (53%) finished with a margin of 5 goals or more

  • 2024 saw the highest proportion (19%) of close games since 1928, and the 4th highest proportion of all time. It also saw 45% of games finish as a blowout, which was the highest since 2018

Percentage of Points Scored in Each Quarter by Era

How has point-scoring typically spread out across the four quarters of games across V/AFL history? Do teams start fresh but end tired leading to more scoring in Q1 and less in Q4, or is the pattern different from that?

The chart below uses violin plots to show, within each era, the distribution of percentages for each of the four quarters. The middle green line on each violin marks the 50th quantile (aka the median).

The median for each quarter and for each era is around 25%, suggesting that attacking and defensive abilities atrophy at about the same rate as matches progress.

One thing that’s interesting to note is the wide spread of percentages, with up to 65% and in some cases 0% of points in a single game being registered in one of the quarters. That spread was greatest in the earliest era.