UPDATE: Close Games in VFL/AFL History: Do Successful Teams Win Them?

Back in November of 2015, I wrote this piece on the topic of close games in response to the claims being made then about the ability of good teams to win close games.

This year so far (2017 to the end of Round 16) we've been spoiled with the number of close finishes we've had, which I'll define now, as I did then, as games that are decided by a goal or less.

We can see, in historical terms, just how high is this year's proportion of close games by glancing at the chart at right. 

Most seasons in V/AFL history have produced close games at a rate no higher than about 15%, and the tendency in recent seasons has been for the proportion to be well below that. This year we're at just under 18%.

The most recent season finishing with a higher proportion was 1976 when it finished at 18.1%, and there have been only 19 previous seasons where the proportion has finished above 17.5%, 12 of those in seasons before the end of Word War II.

In the remainder of the analysis we'll be looking at team results by era. For today's analysis I'm going to use longer eras than I did last time because larger samples of games for each team reduce the associated sampling variability. Across five year eras a team will play about 100 games (fewer in the early VFL seasons); across twenty years, which is the era size I'm using now, they'll play about 300 to 400 depending on the era.

The chart below shows the proportion of close wins and close losses for each team in every era, where draws are counted as half a close win and half a close loss. Each cell is shaded based on the total proportion of close games that a team played in during the era, the overall shading pattern revealing the general decrease in the proportion of such games in more recent times, notwithstanding the spike this season.

The numbers in each cell provide the percentage of close wins and close losses that a team has recorded in a given era. The fact that these numbers rarely differ by more than a couple of percentage points tells us that no team has had a prolonged period of being "good" or "bad" at winning close games.

We can see, however, from the shading patterns within eras, that some teams have been more prone to finding themselves in close games during those seasons. For example, in the modern era, about 16% of Sydney's and Geelong's games have been close, but only 9% of the Brisbane Lions' and 10% of Adelaide's and Fremantle's have been the same.

During this era, the Brisbane Lions, Adelaide and Sydney have had the worst net records, winning only about 40% of the close games they've been a part of. No team has a better than 59% record, and only Geelong, Essendon, Fremantle, the Western Bulldogs and Hawthorn have managed better than 55%.

Based on those results it would be hard to rule out the hypothesis that close games are mostly coin tosses (or, a bit more subtly, that teams win or loss close games more often than their pre-game estimated probabilities would suggest, since not all teams have been of the same average ability across this era).

The best all-time record belongs to the Brisbane Bears, who won almost 59% of all the close games they played in. Essendon is next best at just under 56%, narrowly ahead of Port Adelaide.

(Note that I'm treating the Lions, Bears and Fitzroy separately for the purposes of today's analysis, as well as Footscray and the Western Bulldogs, and Sydney and South Melbourne.)

The worst all-time record belongs to University, who won only one-third of their close games. The Lions aren't far ahead of them though, they having won just 37% of their close games. Adelaide (43%) and West Coast (44%) also have notably low all-time figures. These lower-than-average win rates might also be perfectly well-explained by the slightly below-average average ability of these teams across the entirety of their existence.

Proportionately, all-time, South Melbourne have played in more close games than any other team (almost 17%), Carlton the second-most (about 16%), and Footscray third-most (15%). Amongst the teams that have played at least 250 games - which excludes the Brisbane Bears, GWS, the Gold Coast and University - the Brisbane Lions have seen their games finish as close games least often (about 10%), followed by Fremantle and Adelaide.

We can also look at each team's Close Win and Close Loss histories separately by era, which is what's done in the charts that follow (both of which are clickable to access larger versions).

We can see from these charts, for example, the relatively large proportion of Close Wins that Geelong have produced in the most-recent era by comparison with, say, Adelaide, GWS and the Gold Coast, but also the relatively large proportion of Close Losses that Sydney has endured in comparison to Fremantle and the Gold Coast. This pattern is consistent with the data in the previous table showing Geelong and Sydney as playing a large proportion of close games, and Fremantle and the Gold Coast as playing smaller proportions. 

Overlaid on both charts are the counts of Premiership and Runners-Up finishes by teams in each era. No strong pattern emerges to suggest that Grand Finalists tend to be involved in large or small proportions of Close Wins or Close Losses.

A more relevant metric might be the one we looked at earlier for just the modern era and all-time, which was the win rate of teams in close games. In the chart below we look at that for all teams in every era.

Analysing this data we find that:

  • The correlation between the rate at which a team has won the close games it's played in during an era and the number of Premierships it collected in that era is +0.21
  • The correlation between the rate at which a team has won the close games it's played in during an era and the number of Runner Up finishes it had in that era is +0.10

That's weak evidence, at best, for a link between ultimate premiership success and an ability to win the close games.

Here's the alternative hypothesis:, that we also investigated in the previous blog: close games are largely lotteries and the rate at which generally successful and unsuccessful teams win them will be very similar. As such, they don't provide much information about a team's ability. What will differentiate teams is the rate at which they win games that aren't close. 

To test this, let's use a new metric, as we also did in the earlier blog, which is the difference between a team's winning rate in close games and its winning rate in non-close games, which are those games decided by at least 7 points. Below is the updated chart for this metric. 

A casual scan suggests that a lot of the green cells, which are for teams winning close games at a higher rate than non-close games, contain no Premierships or Runner-Up finishes, and that the red cells tend to contain many Premierships and Runner-Up finishes. An analysis of the underlying data supports this observation:

  • The correlation between teams' winning rate in close games less winning rate in non-close games in an era and the number of Premierships it collected in that era is -0.50
  • The correlation between teams' winning rate in close games less winning rate in non-close games in an era and the number of Runner-Up finishes in that era is -0.55
  • The 49% of teams that have recorded a negative difference between their winning rate in close games and their winning rate in non-close games for an era have won 83% on Premierships and collected 81% of the Runners-Up medallions in matched eras.

Looking at winning rates in non-close games alone provides the strongest evidence of all, with the correlation between a team's non-close game winning rates in an era and the number of Premierships they collected +0.67, and between this and the number of Runner-Up finishes in the era +0.64.

Our alternative hypothesis seems to be supported by the evidence from this updated analysis too. Successful teams don't so much tend to win the close ones, but instead tend to win the ones that aren't close.