Draft Pick Odds

[matchesmalone]

I have compiled a spreadsheet of every draft pick between 1990 and 2006 based simply on whether each pick was used successfully to select a regular NHL player. Here a regular NHL player was defined as either a skater who played at least 400 NHL games, or at least 300 NHL games and three full seasons in the NHL (at least 60 games), OR a goalie who played at least 150 NHL games or at least 100 NHL games and three full seasons (at least 20 games would suffice for a full season as a backup). There were also a small handful of players who I gave an exception - the only two I can remember off hand were Pasi Nurminem, a goalie who played 125 career games, but two of them were full seasons as a starter of 52 and 64 games, and Robert Nilsson, a center who played 252 NHL games and three full seasons, but one of them was a season of 41 points in 71 games. So as you can see, they were fairly impact players when they did play.

OK, so on average, 41.75 regular NHL players - according to my definition - come out of each draft. I felt quite rectified when I noticed that the most players (50) came out of the 2003 draft - generally regarded as one of the deepest drafts of all time; and that the least players (29) came out of the 1999 draft - generally regarded as one of the worst of all time.

Now to the good stuff:

On average,

- a first round pick (1-31) yields an NHL player 55.2% of the time
- a second round pick (32-62) yields an NHL player 23.8% of the time
- a third round pick (63-93) yields an NHL player 17.5% of the time
- a fourth round pick (94-124) yields an NHL player 10.3% of the time
- a fifth round pick (125-155) yields an NHL player 9.7% of the time
- a sixth round pick (156-186) yields an NHL player 10.7% of the time
- a seventh round pick (187-217) yields an NHL player 7.5% of the time

Note that I stopped gathering data after pick 217. The NHL draft only have 217 picks right now so I decided anything beyond that, while perhaps interesting, wouldn't be particularly useful.

To break things down further:

- a top 10 pick: 75%
- a top 20 pick: 63.75%
- a top 40 pick: 49.0625%
- a top 50 pick: 44.5%

- picks between 11-20: 52.5%
- picks between 21-31: 39.7727%
- picks between 32-41: 26.875%
- picks between 42-51: 25.625%
- picks between 52-62: 19.3181%

Success rates by pick for the first round:

- first overall: 100%
- second overall: 100%
- third overall: 87.5%
- fourth overall: 81.25%
- fifth overall: 87.5%
- sixth overall: 62.5%
- seventh overall: 62.5%
- eighth overall: 68.75%
- 9th overall: 50%
- 10th overall: 50%
- 11th overall: 68.75%
- 12th overall: 62.5%
- 13th overall: 68.75%
- 14th overall: 62.5%
- 15th overall: 25%
- 16th overall: 37.5%
- 17th overall: 50%
- 18th overall: 43.75%
- 19th overall: 50%
- 20th overall: 56.25%
- 21st overall: 62.5%
- 22nd overall: 31.25%
- 23rd overall: 50%
- 24th overall: 43.75%
- 25th overall: 50%
- 26th overall: 43.75%
- 27th overall: 43.75%
- 28th overall: 31.25%
- 29th overall: 43.75%
- 30th overall: 25%
- 31st overall: 12.5%

[Josh]

Wow. Thanks for doing this man. There is so much here to work with. Some really cool things that stand out right away:

Sixth round picks tend to yield more NHL players than fourth or fifth round picks.

29th overall tends to yield more NHL players than 15th overall.

21st overall tends to yield more NHL players than 9th overall, and as many as 6th overall.

[matchesmalone]

Yeah the first round is all over the place. Pinpointing specific picks would seem more like superstition than sound reason, so it would probably be more useful to break them down into 3-5 pick segments, to build a larger sample size into each datum. The in-between method may be more handy than the very specific single pick or the very broad breakdown by round.

1-5: 91.25%
6-10: 58.75%
11-15: 57.50%
16-20: 47.50%
21-25: 47.50%
26-30: 37.50%
31-35: 25%
36-40: 27.5%
41-45: 28.75%
46-50: 23.75%
51-55: 21.25%
56-60: 16.25%
61-65: 21.25%
66-70: 18.75%
71-75: 20%
76-80: 16.25%
81-85: 13.75%
86-90: 18.75%
91-95: 13.75%
96-100: 12.5%
101-105: 10%
106-110: 11.25%
111-115: 5%
116-120: 8.75%
121-125: 10%
126-130: 11.25%
131-135: 15%
136-140: 8.75%

And from there it mostly fluctuates around 5-10%, gravitating more toward 5% in the 7th round.

Note dropoffs after the top 5, the top 15, and then less and less significant dropoffs after the end of the first round. But then in relative terms, there is a noticeable tendency to dropoff after around 75, and then again maybe around 90 or 95.

[Josh]

Here's what I think is the most sensible way to break down the first round given the data:

1st - 2nd = 100%
3rd - 5th = 85%
6th - 14th = 61%
15th - 29th = 44%
30th = 25%
31st = 12.5%

I suspect that the 31st pick will become much more likely to yield NHL players in the future simply by virtue of it now being the last pick of the first round rather than the first pick of the "rest of the draft".

[matchesmalone]

Yeah I like that breakdown, at least up to 29. But it just doesn't seem right to isolate 30 and 31; there isn't a real dropoff, because there's a bounceback immediately after. It has to be more of a coincidence than anything else... But then there also could be some rational explanation that I'm missing. The first pick of the second round is a bad team and tends to have bad scouting? Or the pick is fine but they tend to have bad development and rush the player? But then thrre's the fact that at least half of the drafts I considered were from the early 90s, when there were still only 20-26 teams in the league...

Anyways, just for fun, how about putting this to some concrete use? Lets consider the Sens picks from this past draft. But again, cherry picking individual draft slots is problematic, so along with that, I'll take each pick as part of a 5-pick cluster going 2 in each direction:

28th overall: 31.25%
picks 26-30: 37.5%

47th overall: 37.5%
picks 45-49: 26.25%

121st overall: 6.25%
picks 119-123: 7.5%

183rd overall: 18.75%
picks 181-185: 6.25%

Haha well, maybe the lucky 183rd overall pick will come through...

Pretty funny though when you consider that the average fan probably expects a first round pick to be at least an NHL player, and maybe a star. 37.5%... and in a mediocre draft class like this year's my guess would be it's probably closer to 30%

[matchesmalone]

K I'm finally getting started on part II today. I know I've said that before, but I've got nothing else to do for the next four hours, got some brews, and I'm sitting down to work.

I've decided that rather than replacing the original, I want this to be a supplement to the original. I know we had some debate over how many games counts as a successful pick, but I just wanna go with the 400 games from the original work.

Honestly, that seems too much to me at this point. That requires five full seasons of 80 games. I feel like a guy with four full seasons would be a pretty solid pick for the 7th round...

But then, the reason I've kept putting this off, is I just could never seem to figure out the best way to go about this. I want to use the original data as a starting point, just ignore the "unsuccessful" picks for now, and then for every successful pick we can say there's an x percent chance that player will be an all-star, the average x number of games, points, etc.

I dunno, I'm just gonna worry about data collection first, and I think it will come together. We can always go back and make adjustments, even to the original list, if we want.

One important thing is just that we have the original data stored safely and backed up. My laptop is ready to fall apart soon, but as long as you have it Josh, and it should be in both of our email histories.

[matchesmalone]

You know what? Depending how drunk I get while I'm working on this, I might start going back to the original list and adding guys with 300 games or some shit...

[matchesmalone]

Hmmmm. David Hale played parts of 7 seasons. Highs of 65, 54 and 58 games, for a total of 327...

That's an average of 42 games a year. That's a legitimate part of an NHL roster, no?

[matchesmalone]

47 a year actually.

Ah fuck it I'll stick with the 400 for now. If he was a half-season player for ten years he'd be good to go. Seven wont cut it.

[matchesmalone]

I'm gonna give it to Brad Winchester. 390 games. One season of 76, one of 67, two of 64. Also played seven seasons, but averaged 57 games.

[matchesmalone]

Oh right. Winchester actually fits my original criteria - OR 300 games and at least 3 full seasons.

I'm having second thoughts about the second clause for goalies though. Mikael Telqvist would be a successful pick...

[Josh]

K I'm finally getting started on part II today. I know I've said that before, but I've got nothing else to do for the next four hours, got some brews, and I'm sitting down to work.

I've decided that rather than replacing the original, I want this to be a supplement to the original. I know we had some debate over how many games counts as a successful pick, but I just wanna go with the 400 games from the original work.

Honestly, that seems too much to me at this point. That requires five full seasons of 80 games. I feel like a guy with four full seasons would be a pretty solid pick for the 7th round...

But then, the reason I've kept putting this off, is I just could never seem to figure out the best way to go about this. I want to use the original data as a starting point, just ignore the "unsuccessful" picks for now, and then for every successful pick we can say there's an x percent chance that player will be an all-star, the average x number of games, points, etc.

I dunno, I'm just gonna worry about data collection first, and I think it will come together. We can always go back and make adjustments, even to the original list, if we want.

One important thing is just that we have the original data stored safely and backed up. My laptop is ready to fall apart soon, but as long as you have it Josh, and it should be in both of our email histories.

I've got '90 to '05 :)

I like 250 games for skaters. Not sure what I said before. Especially since now we're digging deeper into how good of picks they are, I think we can afford to call a 250-game player a basic "successful" pick.

[Josh]

Hmmmm. David Hale played parts of 7 seasons. Highs of 65, 54 and 58 games, for a total of 327...

That's an average of 42 games a year. That's a legitimate part of an NHL roster, no?

In my opinion, definitely.

[Josh]

Oh right. Winchester actually fits my original criteria - OR 300 games and at least 3 full seasons.

I'm having second thoughts about the second clause for goalies though. Mikael Telqvist would be a successful pick...

Make a full backup season 25 games for a goalie and Telqvist no longer cuts it. Who else would get cut off though?

[matchesmalone]

I was thinking the same thing about 25 games for goalies, but I'm on the fence. I mean, he was a proper backup for four full seasons. That's gotta be as valuable as David Hale, no? Hale was a bit piece, played a depth role for half the games. Bottom pair defenseman when he did play. As backup goalie, Tellqvist was half (ok, a quarter based on GP) of the team's goaltending tandem. When he does start, he's in it for the full 60 minutes, everything riding on him.

I dunno, I dont really mean to take a stance on any of these right now, we just need to talk this stuff out.

The funny thing is, we've kinda flip flopped. Whenever we were talking about this last time, you were thinking of a higher limit and I was thinking lower. I actually kinda agree about maybe something around 250 games, but I'm undecided. For the time being I'm thinking I'm just going with the limits from the original project, and we can always go back (to both lists) and make adjustments. I feel like once we get a decent little sample of maybe three drafts done we'll have a better idea what we're doing.

The only two players from the 2000 draft who are up in the air are Hale and Tellqvist. Even a guy like Andy Hilbert with 307 games makes it based on the three full season rule. And Marcel Hossa would still miss a 250 game cutoff, with 237.