The AFL introduced for the 2015 draft a ‘points value index’ which essentially aimed to ensure that clubs whom have father-son or academy players were paying a price that was a set percentage discount from fair value. This was put into works after multiple young stars whom would normally have cost other clubs a much higher pick were selected in previous drafts, such as Isaac Heeney at pick 18 and Joe Daniher at pick 10 due to the previous regime of a club having to match the bid of another club with their next available pick. Hence despite Heeney being touted as a clear top 3 pick in his draft year and consequently was bid on by Melbourne at pick 2, the Swans only had to use pick 18 on him as that was their next available pick.
This system also has given clubs a clear insight into exactly how valuable the picks they hold actually are, or at least how valuable the AFL deem them to be. Their exact process in determining this was using player salary data from the past fourteen seasons, thereby essentially using salary as a proxy for the quality of player, a very reasonable assumption. Given this is not publically available data, the best that can be done by an outsider is using games played as a proxy for performance and then logically deduce how the graph would shift given other variables being included to better approximate the value of a specific draft pick. Using a points system such as Supercoach which is specifically designed to assess player performance could potentially be a better proxy, however comes with its own issues. Key defenders and to a lesser extent, key forwards tend to underperform in this system. Also whether to use total Supercoach points over a picks career, their peak Supercoach average or some other variation is another tricky decision and so the simplicity of using games played as the proxy for now was deemed more appropriate.
To analyse this, the games played by each draft pick from the 1995 to 2004 drafts were collected. This sample was decided due to 2005 and onwards having at minimum 15 currently active players selected in each draft and thus still are very much ‘in progress’. Drafts pre-1995 were omitted as the further the drafts are from the current, the less relevant the data due to the stark differences in clubs expenditure, time and effort placed into recruiting between now and 30 years ago. Thereby this 10 year sample seemed the most appropriate for this analysis.
The above graph was designed by calculating the percentage of players selected after a draftee whom then went on to play less games. This is much more fair than calculating this for the draft as a whole, as including players not available at this pick is not relevant to the clubs decision at the time due to not actually being able to select said players. It shows a seemingly logarithmic trend around a majorly erratic sample of data, whereby there is only a difference of roughly 5% between pick 14 and 35 via the trend line. Another finding is that the certainty around a draftee playing more games than draftees selected later is much higher around the top 3 picks than elsewhere, as depicted with the standard deviation of this statistic for each pick below.
However this does not say anything about the actual number of games expected at and between these picks, so on its own the graph says nothing about the actual output, only relative output. With the below graph, the median expected games played of a pick 14 has been approximately 90 whilst for pick 35 it has been approximately 50 via the logarithmic trend line. Thereby the small difference in the chances of later players playing less games is actually based around very different expected values and thereby the difference between these picks is still definitely very relevant valuation wise. It may however shed light on the rationale behind Hawthorn’s strategy of trading their late to mid first round picks for established players to extend their premiership window. The uncertainty around their picks was so great that trading for a known entity was actually the most logical thing to do (that is, assuming other clubs were unaware of this themselves).
The logarithmic approximation being the most relevant in our analysis bodes well for the validity of the AFL’s draft pick value index. Below is both together on a graph, with the AFL valuation scaled appropriately, that is that the AFL’s index is scaled down at each pick so that the value for pick 1 is equivalent for both due to it being the yardstick by which both systems are constructed. It is evident that the median games played method values later picks relatively higher than the AFL’s index. This could easily be due to the downside of using games played as a proxy for performance, that even though all else equal you would expect a player with higher games played to be a better player, this is not inextricably tied into a players performance being better. There would numerous examples of this within one draft, let alone across multiple years. Another downside of my method is that, all else equal, you would expect a player picked up at a later draft pick to perform worse than one picked earlier. Hence given salary is tied in practically directly to the expected performance of a player, you’d expect the AFL’s index to attach a lower value as pick number increases relative to the median games played trend line. Thereby taking into account salary would definitely help better explain the variation in performance, holding games played constant.
Another factor that the median games played trend line doesn’t account for is that list spots are limited, therefore even though later picks become less and less different value wise, simply stockpiling late picks is not a viable strategy. This makes them less valuable relatively, thus placing value on pick 80, especially since at that point most clubs would have stopped selecting and therefore you have unlimited picks to yourself, makes little sense and would likely be abused by clubs in attempts to gain multiple small amounts of points for their father-son or academy prospects.
Thereby it seems the AFLs draft pick value index is likely pretty much spot on in relation to how the draft picks should be valued by clubs given a rather elementary but useful analysis seems to show great similarity to their system whilst the differences can be explained in a very logical straightforward manner.