The value of a draft pick

As the Avs started sinking in the standings I started wondering what the value of going from a top 15 pick to a top 3 pick was for the Avs. Then the trade deadline happened and I started wondering exactly what the value of a draft pick, any draft pick, really was. You see a guy traded for a 7th rounder and have to think that even if he plays only 5 games he’s worth more than a 7th rounder.  Then a twitter follower tried telling me that a 3rd round pick was close to a 2nd round pick, and that Burke’s offer for Liles wasn’t so bad. He’s wrong, bt I wanted to know how wrong.

So I decided to collect some data on the value of draft picks to try and compare how valuable different picks are in the draft. The first step was to try and determine how to measure a players “value” which is something that can be incredibly hard to quantify, especially if you start taking stuff like the scoring effects between different eras into effect. I went ahead and started with this approximation:

Or Player Value = (Career TOI/g) / (Position average TOI/g)

I decided that a player who plays more would, on most occasions, be a better player than one who played little minutes. So I decided to strip everything away other than Time on Ice. As I will point out later, I think there are a series of refinements that can be made to gain an even more accurate understanding, and there are some obvious problems from the get-go , for example Goalies. For skaters, however, I believed this is an excellent first order approximation, and I think the results below will confirm that initial belief.

I believe Tom Awad’s GVT vs Draft pick would provide the best analysis, but I cannot find the data, and I don’t understand GVT well enough yet to feel comfortable using it. However, I think TOI/g is a very good 1st order approximation.

After starting with this definition of player value, I went ahead and collected the Average time on ice for every player drafted between the 1997 NHL Entry draft and the 2006 NHL entry draft, for player who played more than 20 career games, and at least 15 games in one NHL season. Any player who played less than that, or no NHL games, was given a 0, otherwise worthless to an NHL club.

I then averaged the value of players drafted in similar positions of the draft  (Top-3 pick overall, 4-6, 7-10… 5th round, 6th round… etc) to get an average draft pick value. I have all that data available on a Google Docs spreadsheet here. I then graphed that average value against the draft pick position (Full Size graph).

</col> </col>
Value Draft Slot
1.23 Top 3
0.93 4->6
0.87 7->10
0.65 11->15
0.70 16->20
0.74 21->25
0.51 25->30
0.35 31->35
0.40 36->40
0.35 41->50
0.30 51-> 60
0.29 61-> 70
0.25 71->80
0.22 81->90
0.24 91->100
0.17 101->110
0.11 111->120
0.17 121->135
0.11 131->150
0.13 151->165
0.13 166->180
0.11 181->210
0.16 211->240
0.12 240+

Please remember. These are all approximations. it’s entirely too early to determine “true” values of players and draft picks. This is an estimation to figure out how valuable draft picks are in relation to each other and NHL players. a .75 isn’t necessarily actually a better player than a .71. They are just used on the ice slightly more in their career.

RESULTS: As expected a top-3 pick is incredibly valuable, and is the only slot which will, on average, produce an above average NHL player. (The most valuable player, Alexander Ovechkin is a 1.49 and any player who didn’t make the NHL is a 0. This will push averages down significantly)

A Mid to late first round pick is almost double value of a second round pick. It looks as if there are ~25  players in each draft with a high probability of being an NHL caliber player, and 2-6 who are head and shoulders above the rest in any given draft. Anything after the first round starts losing value quickly.

Any draft pick after the third round is essentially the same. In fact shrewd GMs will try and trade away 4th round picks for multiple lower picks since they all have about the same probability of producing NHL players.

Any trade in which a team acquires a serviceable NHL player for a 4+ draft pick is automatically a steal. I.E: The Daniel Winnik for a 4th rounder is highway robbery.

The Burke Kessel trade was a nightmare. Burke got Kessel (Value: 1.14) for a top-3 (1.29), high 2nd round (.35), and this years 1st rounder (currently at 8th: .87). Now these values are by no means concrete and they aren’t true values of any players, but only approximations based on how much players play. But no matter how big the error bars are on these values, Burke straight up got robbed.
And here’s the most important point: I was right. Liles is worth a shit-ton more than a 3-rd round pick.  Jon Michael Liles (Value: .99) is worth well more than a 3rd round pick (.29).  Brian Burke got robbed in the Kessel trade and may be trying to overcompensate by trying to rob someone else. A fair starting point for negotiations value would be a second + prospect, or late first rounder for Liles. By the way, there’s no way Burke doesn’t know that, he’s just a jackass.

NOTES:   I’d really like to do this with Tom Awad’s GVT vs Draft slot, instead of TOI/g. (I’d go with Average GVT/season for each player) There’s 2 reason’s I didn’t. The first being I am not familiar enough with GVT to feel comfortable using it. But even more detrimental was that I could not find the data necessary. I’d be interested in seeing it, and I’ve got an e-mail to Tom to see if he has the data for it from 1998-2010.

The NHL only started keeping Time on Ice data in 1998, which is why the data goes back to players chosen in 1997. Before that players would be skewed by only averaging part of their careers. The reason 2006 was chosen was because most players who will have measurable value from that draft class have made the NHL already. It was a balancing act, as not enough years would yield too few data points to make this endeavor worthwhile.

Games played was not taken into effect for precisely this reason as well. There are still many players from the 1997 draft still playing, and nearly all from the 2006 draft still playing. It’s difficult to find a way to measure how valuable GP when so many careers are still going on, and the complexity of the weighting required to make 2006 draft picks equal to 1997 ones is fairly complex (and unknown at this point.) The nice part about TOI/g is that it evens out that gap fairly well. Albeit with some bad data points.

I had some difficulties figuring out how to factor in goalies, so I just deleted them from the data. This was only a few data points, and saved a lot of work, and I think the effect if adding them in would be negligible.

Players killed before reaching the NHL, such as Luc Bourdon and Alexei Cherenepov, were also deleted from the data.