The best AFL player of all time using the most important statistic
How to determine the best AFL Player
A lively debate, the best AFL player can be determined many ways. It also depends on one’s definition of the ‘best’. Is it Tony Lockett, who’s kicked the most goals ever at 1360, or his potentially more well-rounded rival, Jason Dunstall who’s kicked 1254, not to mention the undoubtedly dominant Wayne Carey who played further up the field. Do we look at Dane Swan who averaged nearly 27 disposals a game over his career along with a Brownlow & Premiership, or Gary Ablett Jnr with a mere 25 disposal average but 2 each of Brownlows & Premierships. Perhaps it is Gary’s even more spectacular father who’s freakish talent constantly left opponents dumbfounded.
We could look at the longevity of players who at the end of their careers were playing against many who weren’t even born when they started out – Brent Harvey and Michael Tuck who played 432 and 426 games respectively. We could discuss the versatility and durability of players like Adam Goodes and Matthew Pavlich who could play anywhere on the field, competing vigorously to beat their opponents much more often than not.
None of these discussions are wrong or unimportant, but what they typically lack is a truly quantitative and objective view of the game. What would be better is if we could focus our analysis on the most important statistic of the game, the number of goals kicked.
The most important statistic
During AFL commentary, recall Leigh Matthews saying “The most important statistic is the number of goals you kick” (I can’t find this quote). This made me think. Maybe to find the best player of all time, we need to compare with the objective of the game, team goals. By extension we can also say that the most important statistic is two-fold:
- The number of goals the team kicks in a match
- The number of goals the team concedes in a match
The next step of this analysis is to find a way to link this team statistic to make a player statistic. There are a multitude of ways this could be done. I will share my first attempt. In brief, I have considered how well a team performs (in terms of goals) in the matches the player plays, compared with the matches the player misses (through injury/suspension/omission) throughout their career. This is considered for all eligible players over the life of their careers, to give each player a score, which I’ve called ‘goal impact’.
Who is the best AFL Player
As you can see in the blue bars, the best player according to this analysis is Darren Jarman. He is widely regarded as one of the most skill-full players to have played the game. Here are a few notes about what the numbers mean:
- Goal Impact is comprised of 2 parts: Goal impact = Goals-For Impact – Goals-Against Impact
- A high Goals-For Impact is desirable – player likely impacts their team favourably in attacking (kicking team goals)
- A low Goals-Against Impact is desirable – player likely impacts their team favourably in defence (preventing opposition team goals)
- For each of these stats, the ‘average player’ who has no impact on their team (favourably or unfavourably) will have a score of 0.
- Darren Jarman has the highest Goal Impact score of 0.1491, and the lowest player has score -0.0909. All other 380 eligible players have a score somewhere in between.
- Not only has Jarman achieved the best Goal Impact score overall, he also has the best Goals For Impact score. This suggests he has also been the most impactful attacking player. He is only ranked #25 for goals-against-impact however.
What does this list consider
There were 382 eligible players considered. I’ve made an intuitive judgement as to those that truly can be considered for discussion of the best ever AFL player. This is to avoid those players who happened to have a few good and/or potentially lucky/unlucky seasons.
They all need to have played at least:
- 1 season in the AFL (i.e. in or after 1990)
- 200 VFL/AFL matches
Additionally, each player needs to have played one or more match in a season for that season to count (e.g. if they miss an entire season, it is not included in the analysis)
The ‘goal impact’ score for each match, carefully makes the following considerations:
- Both: team Goals-For and Goals-Against.
- Adjusts (scales) score relative to team’s performance for that season. For example if Adelaide’s median goals in season 2016 is 16, but in one particular game they kick only 12 goals, then this game is considered a bad game for Adelaide, than say if they had kicked 18.
- Adjusts (scales) score relative to opposition team’s performance for that season. For example if Adelaide kicks 15 goals in a game against Carlton who’s median goals-against is 17, then this game is considered a bad game for Adelaide, than say if Adelaide had kicked 19.
Why is this different to what we’d expect
Of those players listed in my introduction (widely regarded all-time greats), only Jason Dunstall, made it into the top ten. The standard ways of looking at the best players consider totals and averages for: kicks, marks, handballs, disposals, individual goals etc. More recently we are able to capture rigorous super-coach and AFL fantasy scores which are linked to the quality of possessions. This is a step better, and it can be used for example to predict who will win Brownlow Votes but its not everything. Furthermore, it’s still not closely linked to the objective of the game, kicking goals.
AFL is a team game. For that reason, it’s very difficult to objectively determine which players are contributing most to their team’s success or failure since there are simply too many players and actions that could be measured. I remember some great team advice from one of my coaches during my amateur career: “It doesn’t matter if you don‘t always have the ball, but make sure you’re doing something all the time”. Meaning, there is more to football than getting a lot of the ball, and wowing the audiences. If you are ruthless in your endeavour, you will do whatever is needed to help your team kick goals for your shear desire to win. Below are just a few rarely discussed/measured statistics, which in isolation seem trivial, but when looked at on the team level make a huge difference to kicking goals and winning games. I estimate that some of these (and others) contribute to the above ‘goal impact’ results.
- Constantly moving around the ground to make space for team-mates or draw opposition players
- Getting to a lot of contests
- Chasing and pressuring opponents
- Smothering, blocking, punching the ball
- Having the mental strength to dispose of the ball well in high pressure situations
- Getting around teammates to provide psychological support
- Getting under the skin of dominant opponents (there’s a fine line)
- Getting in the right positions to stop or start play as needed
- Setting an example to team-mates to play smartly, team-first and hard at the ball
- Pushing and encouraging team-mates to give everything they have
- Being an accurate kicker so that your team doesn’t waste opportunities
The best attacking players
Here are the players who impact their team’s Goals-For the most, as shown in green (Goals-For Impact). Other stats, (overall) Goal Impact and Goals-Against Impact are shown in blue and red for reference.
What is interesting about this is that many of the top 10 players, make little impact to their team’s defence (red bars close to 0 line), or in some cases they appear to have a unfavourable impact (red above 0 line). Perhaps if these players were also the elite of the elite in defensive measures, their Goal Impact would be considerably higher in the discussion of ‘best of all time’.
The best defensive players
Similarly, here are the players who impact their team’s Goals-Against for the most shown in red (Goals-For Impact).
Perhaps the most surprising result from this list is Eddie Betts who plays most of his time in the forward line. The remainder are either defenders or midfielders who roam forward. Betts could be regarded as one of the most valuable current players in the competition. His pressure on opponents results in fewer opposition goals scored, despite his defence happening predominately in the forward line. I estimate that Betts’ defensive efforts results in his team having the ball in the forward half for longer, meaning less opportunities for the opponent to score. Hence even forward line players can be extremely valuable to their team stopping goals.
This is one way of moving closer to quantitatively finding the best players of all time. I stress that this is only one approach, and like all other statistics, it is merely that. We are counting certain values to provide insight and estimation of the world through a model. As said by British statistician George Box ‘All models are wrong, some are useful.’ The 3 ‘Goal Impact’ statistics I have created, can be used to supplement other statistics when analysing the game for: best players of all time, player performance, drafting, development, game predictions and so on. What I would recommend however is improve the statistic for greater accuracy and refine it for the purpose needed.
Here are some of the limitations and further consideration of this analysis and hence the 3 statistics
- Some teams are better are replacing players when they are absent from a match (great coaching and depth). For players in these great teams, their impact when playing or not is less felt by the team’s results on the score-board.
- Similarity for some teams that are poor at replacing players when they leave a team (poor coaching and depth).
- Some teams have many favourable/unfavourable players in the team at one time, perhaps conflicting with one another’s statistic.
- Players have peaks and troughs throughout their career. Should we penalise or ignore players having a few sub-par seasons?
- Do we choose only the best 5, 6, 7 (or more) seasons for a certain player and ignore the rest?
- What about players never to have made it to 200 matches. Is 200 that a fair cut-off? (Although, I couldn’t think of any ‘best of all time’ type players in this category).
- Some players have missed very few games that their impact to their team in their absence is hard to measure (e.g. Shane Woewodin, Harry Taylor & Joel Selwood who have missed < 15 games in their 200+ game careers).
- Teams’ performance will peak and through throughout the year. Perhaps we need to consider goals kicked relative to team’s form over say, the last 5 matches (instead of season).
- Add more rigour of the statistical approach to this problem. This is a model-based scoring system considering how players are linked to team goals. But perhaps we could make a model to predict how many goals a team will kick when said player is in, compared to when they are not.