This post acts as an “appendix” to my introduction of our draft pick value chart developed from trade data.
Using trade information from CapFriendly–i.e. the most indispensable site known to hockey fandom–I compiled a list of trades involving only draft picks from the 2008-09 season through today. I filtered the trades on CapFriendly to entry-draft-related trades, since this is when almost all such trades would occur. Then, if a trade involved a player or a player’s signing rights I discarded the trade from my sample because I wouldn’t be able to compare the value of the player to the value of the picks swapped. This yielded a total of 184 pick-for-pick trades.
From that list I further narrowed the data to include only those trades of picks within a single draft year. I did this for two reasons. First, it is intuitive that a draft pick this year would have more value than the same draft pick next year; the team potentially receives the benefit of the player sooner. This is a basic time value concept. Therefore, when building a draft pick value chart, trades for future picks would need to be discounted. But how much? We can’t answer that in the abstract, so future-pick trades could introduce some bias.
Second, and more practically, the actual draft pick position of a future pick is ordinarily unknown at the time of the trade. Seattle may know it is acquiring Florida’s 2023 sixth-round pick in a hypothetical deal, but what pick would that be exactly?
For these reasons, when constructing the basic trade value chart, I used exclusively draft-pick-only trades made within one season. There have been 98 such trades since the 2008-09 season.
This information was then processed through various functions to determine best fit and through a maximum likelihood analysis, followed numerous spot checks, and a few manual adjustments, as described below. I then normalized the values so that the first overall pick would be worth 1000 “points.”
Key limitation: Uncertainty at the top
One significant limitation of this study: Since the 2008-09 season there have been zero instances of a top-10 pick being traded in a pure pick-for-pick swap. The highest pick traded was the 11th pick, which happened twice–once in 2016 in a swap for the 12th and 80th picks and once in 2019 in exchange for the 14th and 45th picks. The lack of data in the top 10 undermines the concept of a “descriptive” trade value in that range. Any “value” I can ascribe in that range is an extrapolation–and more or less speculation.
As it turns out, the top-10 values produced by my model were too “flat” in my opinion. They indicated, for example, that the first overall pick could be acquired in exchange for a package of the fifth overall pick and the 36th overall pick. This rang false to me. It seems to me it would take more than a high second rounder to move from five to one.
So, for the top-10 picks, I manually built the projected pick value on a steeper curve commensurate with Dom Luszczyszyn’s GSVA model of the draft. My thought was that the projected quality of the players that high in the draft could serve as a reasonable proxy for what teams could be expected to “pay” to acquire the pick.
Setting aside this admittedly improvised solution for the top-ten, the actual values indicated by my study are reflected below:
I also linearized the decline in draft pick value from approximately the mid-fifth round through the end of the draft to reflect that there was a marginal but real decline in value for each of these picks. My original models flattened or were otherwise less logical by this point of the draft.
Trade value chart performance
This model performed well on the whole. Attempts at manual adjustments over the last few days have typically worsened its performance as a descriptive tool, so I’m confident this is fairly close to the market. The chart yielded an average error and median error of less than one “point” of trade value, with a good deal of the error coming in a small handful of transactions where a team paid an atypically high price to move into the top-half of the first round.
The performance is particularly strong considering the model has no way to adjust for the quality of the specific players or the depth in a given draft. For example, in one year, there may be 35 players that a team considers to be “first round” talent, but there may only be 15 such players the next year. Presumably, a team making such an evaluation would be willing to trade more to move up into a late-first-round pick in the former year rather than the latter year.
A few parting thoughts
1. Few trades overall. It was notable to me that over a span of 13 years, there were only 184 pick-for-pick trades in my sample. NHL teams are fairly conservative and comfortable “sticking and picking” when it comes to the entry draft.
2. No top-of-the-draft pick swaps. As noted above, I was surprised that since the 2008-09 season there had been no pick-for-pick swaps in the top-ten. Going farther back, there are examples of such trades and the returns for trading down were typically exorbitant. Qualitatively, this supports my decision to upwardly adjust the value of top-10 picks in my model. That said, again, the values I have ascribed to trades in the top 10 should be taken with a fair amount of salt.
3. It is costly to move up. The decline in trade value is steep near the top of the draft; this makes it “costly” for teams trying to move up. For example, a move up of just six or seven picks in the high second round could cost a team its third-round pick. Perhaps as a result, there were very few trades in which a team climbed more than a dozen spots in the draft.
4. Overall trade behavior may not have changed very much since 2006. This is something I want to look at more closely, but the trade value chart I discerned is not entirely different from the chart developed by Eric Tulsky looking at trades from 2006-12. This chart is a little steeper, i.e. more costly to move up, but it is not different in kind. This is just one data point, but it suggests to me that there has not been a seismic change in how teams look at the value of draft picks over the last 15 years.