One of an NHL general manager's most important resources is the draft choices he receives at each year's Entry Draft. The trading of draft picks is very common, and sometimes draft picks alone are used to acquire players. In September of 2009, Toronto GM Brian Burke traded away his team's first- and second-round choices in 2010, and first-round choice in 2011, to acquire Phil Kessel from the Bruins. How does one go about determining whether a trade like this makes sense?
Obviously, one needs to project the future value of the resources that are exchanged by the two teams. Kessel had three NHL seasons under his belt at the time of the trade, recording the following GVT values:
Phil Kessel, ages 19-21
Age GP GVT GVT/82
19 70 4.0 4.7
20 82 3.9 3.9
21 70 14.7 17.2
This value pattern is not at all unusual. In his first two seasons, Kessel was a useful player with very similar GVT rates. In his third season, he was given more offensive responsibility and performed very well, increasing his GVT rate to a very high level. Looking for players with a relatively similar value pattern, we find Patrik Elias and Keith Tkachuk, for example:
Patrik Elias, ages 19-21
Age GP GVT GVT/82
19 77 3.6 3.8
20 75 4.6 5.0
21 77 10.1 10.8
Keith Tkachuk, ages 19-21
Age GP GVT GVT/82
19 17 0.8 3.9
20 81 3.3 3.3
21 82 12.1 12.1
As it happens, both of these players have a peak GVT value of just over 20 (Elias 20.7, Tkachuk 20.4). Peak GVT value is defined as the player's GVT in the best five of 10 seasons after his draft year, on a per-82-game basis. It is therefore not unreasonable to think that Kessel will produce GVT values of around 20 per season when he reaches his peak. But even if we accept this value as a reasonable one, how do we go about valuing the other side of the equation, the draft picks?
It should be a simple matter to estimate the value of these picks. We have many past Entry Drafts to examine, and we know the GVT values produced by the players selected in those drafts. We need the drafts to be relatively recent, to try and match current conditions, but also not so recent that we don't know the drafted players' peak GVT values. The years 1997 through 2001 give us a good sample that fit our requirements well. The method couldn't be easier: simply calculate the peak GVT values for each draft position for each year, and then add the values together for each draft position. Note that we use the draft-ranking version of peak GVT (that is, defensemen have their values multiplied by 0.8 and goaltenders by 0.33) to avoid any confusion caused by a player's position.
For example, the first overall draft picks from 1997 to 2001 are:
Year Name Pos GP GVT GVT82 GVT82+
1997 Thornton, Joe F 389 103.4 21.8 21.8
1998 Lecavalier, Vincent F 404 93.8 19.0 19.0
1999 Stefan, Patrik F 326 30.6 7.7 7.7
2000 DiPietro, Rick G 276 51.0 15.2 5.1
2001 Kovalchuk, Ilya F 407 86.2 17.4 17.4
GVT82 is the player's peak GVT per 82 games, and GVT82+ is that number adjusted for position. The average of these numbers (not a simple average but a weighted average based on games played) is 15.1. That's our first data point. Repeating the process for the first 60 draft picks from 1997 to 2001, we get the following data set:
There's quite a bit of variation due to a relatively small sample size of only five drafts; since we don't want the 15th pick to be valued at less than the 60th, we need to fit a line to this data to arrive at our estimated value for each pick. If you remember your high school math, you might recognize that the basic shape of the line will be in the following form, where y is the estimated value of the draft pick, a is some amount to keep the line from reaching zero GVT too quickly, b is some other number, and x is the draft pick number, or some number times the draft pick number:
y = a + (1 / b) ^ x
Sometimes an -inator can be a simple function like this, arrived at by fitting a curve to some data. So here is the Valuatinator, which you can use to estimate the peak GVT value of a draft pick, where v is the estimated value of the pick and d is the number of the pick (1 for the first overall pick, 23 for the 23rd overall pick, etc):
v = (4.1 x (200 d) / 210) + (20 x .7 ^ .9d)
This function gives us values such as these:
Draft Pick Peak GVT
This is a dramatic illustration of the drop-off in value from the first few picks in the draft to all of the other picks. Starting around pick number 10, the reduction in value is very gradual, finally reaching an estimated peak GVT of zero around pick 200. The following graph shows the function fitted against the data:
Now we have values for each draft pick in an Entry Draft. To return to our analysis of the Kessel deal, we now know that the second overall pick and the 32nd overall pick in the 2010 draft are worth a combined peak GVT of 17.7 (14.4 plus 3.3). However, there's still the 2011 first-rounder to consider. Its value obviously depends on where the Leafs finish this season. If the season were to end today, Toronto would probably pick around number five, which would be worth about 7.8 peak GVT, which in turn would make the total peak given to acquire Kessel 25.5 versus Kessel's value of 20.0.
But it's not that simple. When Burke made the deal, he did not know where the traded picks would end up being in the draft. That the 2010 first-rounder would be the second overall pick was not known at the time, and that's the danger of trading your picks before you know where those picks are going to be. If Burke convinced himself that Kessel would be the player needed to make his team a playoff contender, that could have been enough to justify the trade by itself. If the picks he gave up were in the middle of a round, he'd only be giving up about 10 GVT in peak value to acquire 20 GVT in Kessel. But he risked giving up something huge. If Toronto stumbles a bit the remainder of this season, he could end up having traded away another second-overall pick, which would make the deal look very bad indeed.
Evaluating trades in never easy, but hopefully the new Valuatinator tool will make it a bit easier.