Surely you realize that even if data is 100% distributed at random, there will still be a "result" where someone will finish in "last" and someone will finish in "first," yes? The fact that Vancouver finished last in no way argues against the idea that the data is essentially random noise. I have never won the lottery whereas I read in the news that some guy recently won it for a second time, but my "grizzly results" that put me in last place in lottery winnings doesn't actually mean a thing.
For the purposes of measuring expected value, let us define a "success" as a player who plays 250 NHL games. I am picking this number because we have used 100 games in the past and others have felt that it was too low a bar. I have tried this analysis with several criteria though (including ice time, points, etc.) and it makes little appreciable difference. So let's use 250 NHL games this time. Feel free to suggest something else if you wish. Now, here is the % of picks who make it to 250 NHL games by pick:
Note: I am removing goalies from this.
You can fit this curve pretty nicely to get a formula which will calculate a rough % of "success" at each pick, which you can then use to calculate the expected number of "hits" for each NHL team based on where they are selecting. Obviously this isn't perfect because the players are different each draft and there's no such thing as a perfect model, yadda yadda yadda.
For the period of 2008-2013 these are the results:
Team | Picks | xSuccess | Success | +/- |
OTT | 40 | 8 | 10 | 2 |
ANA | 41 | 9 | 11 | 2 |
WSH | 36 | 7 | 8 | 1 |
MIN | 33 | 7 | 8 | 1 |
NYI | 42 | 10 | 11 | 1 |
SJS | 37 | 6 | 6 | 0 |
TBL | 37 | 8 | 8 | 0 |
CBJ | 38 | 8 | 8 | 0 |
NYR | 33 | 7 | 6 | -1 |
LAK | 40 | 8 | 7 | -1 |
NSH | 43 | 8 | 7 | -1 |
WPG | 19 | 4 | 3 | -1 |
BUF | 44 | 9 | 8 | -1 |
NJD | 36 | 7 | 5 | -2 |
ATL | 22 | 5 | 3 | -2 |
PHI | 30 | 5 | 3 | -2 |
BOS | 33 | 6 | 4 | -2 |
CGY | 36 | 7 | 5 | -2 |
DET | 40 | 7 | 4 | -3 |
COL | 34 | 8 | 5 | -3 |
CHI | 47 | 9 | 6 | -3 |
VAN | 33 | 6 | 3 | -3 |
STL | 38 | 8 | 5 | -3 |
CAR | 33 | 7 | 4 | -3 |
FLA | 45 | 10 | 7 | -3 |
TOR | 39 | 8 | 4 | -4 |
EDM | 45 | 11 | 7 | -4 |
PHX | 36 | 8 | 4 | -4 |
DAL | 35 | 7 | 3 | -4 |
MTL | 37 | 7 | 3 | -4 |
PIT | 32 | 6 | 1 | -5 |
[TBODY]
[/TBODY]
So here we have the Canucks with ~6 expected players and they in fact got 3 (Hodgson+Connauton+Horvat.) If you don't like my criteria then you are free to suggest different criteria and I will do it again. I realize that Hodgson is not generally considered a "hit" so I am more than happy to run this in various ways based on whatever criteria we want to agree upon, but I really don't think it makes much difference to my overall point. The reason why the results are skewed towards the minuses, by the way, is because there are players in the recent drafts who have not made it yet to 250 GP but likely will (e.g. Ben Hutton,) so you can probably add 1 to each team and you get something closer to what the ultimate result will be.
I should clarify a few things because sometimes, especially when posting from my phone, I can speak in strong and dismissive ways which is to my own detriment when someone intelligent responds. When I say "it's all random" what I mean is "I have not seen anything convincing that suggests it is anything but random (i.e. to suggest there is 'skill.')" And when I say "there is no skill" what I really mean is "there is no observable difference in skill between different NHL GM's based upon the data we have."
No matter what method you use, including your very own analysis that you posted (thank you by the way,) there is nothing in any of the results that IMO cannot be explained by simple random variance. I believe that if you were to distribute the results randomly, you would see a similar distribution as to what we actually observe here. The "top" teams got a couple more players than expected based on where they were selecting, the bottom teams got a couple fewer, and the middle teams got more or less the amount we would expect. I could probably prove this more convincingly with a simulation, but the
potato illustrates this fairly decently as well. Simply drafting players based on a rudimentary linear equation does not yield worse results than the majority of NHL teams. This suggests to me that whatever "skill" is involved does not vary strongly between teams and is highly overstated.
There are a handful of "reasonable" selections at each spot and whether or not you get a "hit" is basically luck. If there was a significant amount of skill to it beyond that, I would expect teams to be able to beat the potato on a consistent basis, such that the results of the potato are laughable and easily the worst results when compared to the results of NHL teams. This is not the case.