With the 2019 AFL season rapidly falling from sight, the attention of the league’s media and fans turns to who will be at each club in 2020 and beyond.
The career HPN Trade Calculator has now been updated for 2019 values, so go and have a look!
Also, we’ve introduced a 2020 Trade Calculator – predicting only one year into the future. Something for every disgruntled fan!
HPN has been analysing the player movement period since 2014, looking at draft pick values, player values and more. For this season, HPN has revamped our player valuation model, after considering some constructive feedback provided over the last year.
Note – all values in this piece are expressed in Player Approximate Values (PAV).
What was wrong with PAPLEY?
PAPLEY (PAv Player Linear Extrapolated Yield) was pretty accurate at projecting the careers of most players, but struggled with the more exceptional players. The outliers are always the hardest to predict, and PAPLEY was ultimately too conservative. with great players, and slightly undervalued older players.
Over the last 12 months, HPN has tested a variety of projection models before ultimately landing on the approach chosen. It is a hybrid model that predicts both the career arc of players and their projected career length. It is the latter that is harder to project, but as a rule of thumb better players play for longer – a concept HPN stumbled on in 2015.
Generally, for every percent better than the average player at a given age a player is, they will play for about one percent longer across their career. There are exceptions and limitations for this, but it’s handy to remember when thinking about how much a player is worth
This means that good young players are still the most valuable in the system, but great older players are more valuable than ever before.
Who are the most valuable players?
Again, young, good players reign supreme. As a guide, pick 1 is worth about 147 PAV, and every player above 135 is worth closer to pick 1 than pick 2. That means there are 13 players projected as producing as much as the average pick 1, the top 13 in this list..
However, historical data suggests about 20 players currently in the league will produce the average of pick 1 across their career. The gap between that and the values above indicates the uncertainty and difficulty of working out who will be those 20.
Take for example a young Simon Black, pick 31 in 1997. Black played just 9 games in his first year – relatively average ones at that. It would have been foolish to predict that a player would become the dominant force that he ended up being from such a small and innocuous start.
Of the 13 listed above, it is likely that about two-thirds end up being as good as the career of the average pick one, with the majority of the remainder producing at least as much value as a top five pick. There will be, however, players who struggle with injury like Anthony Morabito, and fail to live up to their promise from here.
Why is my favourite player/potential trade bounty rated so poorly?
There are three critical reasons that players are under- or overrated: age, injuries and exposed AFL form.
Let’s take Joe Daniher for example. According to PAPLEY MKII, two years ago Daniher would have been one of the most valuable players in the game – a 23 year old superstar coming off multiple good seasons. He would have been comfortably worth pick one, and potentially more.
Right now, PAPLEY rates the 25 year old Daniher much more unfavorably. After struggling with injuries and form, and seeing two years pass by with little to show for it, PAPLEY believes, based on the dataset of all past players with similar recent output, that Daniher is likely to produce 22 more PAVs across his career. That’s roughly worth pick 60.
*cue screams of outrage from Essendon and Sydney fans*
It is highly likely that Daniher will outperform his current projection, but also quite unlikely that he will live up to his projection from two years ago. It would be irresponsible for a model to overvalue a player based on reputation rather than actual output. It remains to be seen how the clubs in question approach the thorny question of Daniher’s worth and future trajectory, given both his talent and injury cloud.
The trade calculator is ignorant of salary values, for the reason that about 99% of the footy world has no accurate knowledge of all salaries. There are some fair guesses, and some public knowledge, but it is vastly outweighed but incorrect or unverifiable claims.
Likewise, the model does not remotely care about the trade rumours floating around, from award winning journalists and burner accounts alike. The model is an account of how much a player should be worth, not what a radio shock jock thinks that their least favourite club should pay for them.
As a side note, almost all trade rumours with at least some correct information that circulate before the trade period, and the majority that circulate within the period, are attempts to gain leverage from one or more of the parties involved in the deal. Whether it is from one of the clubs in question or a player agent, always try to analyse where the source of the information might be, and how much credence to give it. Don’t buy into the information imbalance.
What’s the error in the predictions?
A little bit! Using nothing but age and season-by-season player values to predict future career trajectories, expressed as a single number, is inevitably going to see errors and outliers.
It’s hard to express one term for error, but generally the model misses more at the extremes, and for younger players, and misses less for more average, and for older players.
For players younger than 21, the predictions for future value should be solid within a span equivalent to give or take a late second round pick (pick 34 or so). At 23, that error spread shrinks to about a mid third rounder (about pick 45), with the error reducing with age.
The model overpredicts slightly, due to projecting future values for some players who will never play again. On a player by player basis, the projections are still relatively accurate.
HPN will try to go into more depth about error rates in the coming weeks.
But who is going to win the trade right now?
Good question! To answer that, HPN has developed a One Year Trade Calculator.
Let’s take Oscar Allen, the young Eagle, as an example. He is predicted as producing about 100 PAV across the rest of his career – good for second at the Eagles. He was good this year, and is very young, a fair bet for having at least a good career. But next year, he is predicted to produce only 10 PAVs – only 13th at West Coast. If a club were to trade for Allen, it would be for his long term promise, not his short term production.
These are the predictions for the top players for the 2020 season:
All draft picks have been given a notional value in both systems – a relatively low one in the 2020 Trade Calculator. This is because draft picks rarely contribute a lot (or at all) immediately, and trading for a draft pick is usually a move for years in advance. If your club is in a premiership window (with more than a 10% chance of a flag for instance), either one of these models will be of more use.