Junior hockey and age, part 2: the NHL draft

Part 1 focused on players at the time they entered the CHL (OHL in particular). Part 2 shifts the focus to the NHL Entry Draft, which just so happens to be less than two weeks away.

Let’s get right into it. The OHL part of this table, you’ve seen before; I took it from the table at the end of part 1, with all the rounds collapsed into one line, so 2,207 players in all. The NHL part is based on new data, including all birthdays of skaters drafted from the CHL between 1998 and 2004, inclusive (N=692).

Birthday distribution of draft selections (non-goalies only)

Draft N Q1 Q2 Q3 Q4 Q1:Q4 Ratio
OHL 2207 44% 30% 16% 10% 4.3
NHL 692 38% 27% 20% 15% 2.5

This is again nothing new and is included here just to establish that the bias, in fact, diminishes once NHL teams are selecting players, though it’s still high. (Later in this post I will point out why it’s too high.) A big difference between the OHL Priority Selection and the NHL Entry Draft is the September cutoff in the latter, and the numbers are higher for Q3 and Q4 birthdays partly because of that, I would imagine.

Time really does move slower on the West Coast

After part 1 was published, Cam Charron wondered which NHL teams were the worst at loading up on Q1 players. I pointed out that “the age thing” doesn’t apply as much once NHL teams are selecting players for more than the next few years, to which he responded:

“I’d wager there are notorious teams who load up on the older, developed players at the draft out of ignorance.”

Well, let’s see.

The average age of a skater drafted from the CHL is 18.8 (which actually means 18 years and 9.5 months — bear with me as I use base-10 for ages). Most teams are centred around that mean. The oldest are Vancouver (19.1), Carolina (19.1), and St. Louis (19.1), each about four months or so older than average. We’ll get back to them in a second.

The teams who “draft young” are Chicago and Columbus. Here’s how extreme the difference is: the oldest CHL skater drafted by the Blackhawks was Troy Brouwer (age 19.1 when drafted). His birthday is August 15th. The average draftee by the Canucks is, as noted above, also 19.1 years old (actually 19.125), which would put his birthday, if he existed, at July 31st.

So Brouwer, the oldest Chicago draftee, is two weeks younger than the average Vancouver draftee.

The obvious question, then: because they took younger players, did Chicago (and Columbus) do a better drafting job than Vancouver and the others?

Results of selected teams’ draft picks (“NHLers” means “played 82 games or more”)

Team N NHLers Average Age GP G A Points per draftee
CHI 24 9 18.4 2751 382 717 45.7
CBJ 20 6 18.5 2055 363 449 40.6
VAN 22 1 19.1 677 42 95 6.2
CAR 24 3 19.1 1292 349 428 32.4
STL 21 4 19.1 1569 93 259 16.7

It would certainly seem so. The Canucks’ CHL picks have not turned out well (points and assists-wise, anyway), nor have the Blues.

Anyway. The “young” teams drafted 44 players, 15 of whom became NHLers; the “older” teams drafted more players and ended up with fewer NHLers.

Yeah, so? Does age matter or what?

Cam also asked if younger players chosen later in the draft were “more likely to surprise.” There are a number of ways to answer this.

For example, if you look at two groups of players drafted late, with similar junior stats, and the only difference is that one group is older and one is younger then we’d absolutely expect the younger players to be better in the NHL. Gabe Desjardins already answered that question, right?

But one can fit a lot of meaning into 140 characters, and you’ll notice Mr. Charron used the word “surprise.” Meaning, did they do better than expected? Not just “did they do well?”

Thankfully we have a way to establish expectations — that very same link to Desjardins’ work shows us the CHL-to-NHL PPG conversions by age. It’s quick and dirty but it works for us.

Applying those conversions to our CHL draftees, and comparing their actual points to their expected points, will allow us to measure in some way how these players over- or under-performed. Granted, PPG is not the best way to evaluate players, but within the context of these conversions it is valid.

We’ll take all the picks in the late rounds. There are 418 of them. We’ll divide them into five buckets based on age and see what they did in the NHL vs. their expectations.

To define outperformance let’s use an example. Former Hamilton Bulldog Michael Ryder was expected to have 16.1 points in the NHL based on his pre-draft CHL stats, his age on the draft cutoff day, and the average number of NHL games played by a 4th round or later draftee (47). However, he has 334 points so far, so his surplus point value is 318.

In the table below, “outperformers” is the number of players whose outperformance was 30 points or better, roughly the 90th percentile for all late draftees. This is the last table in this post, promise.

Outperformance of late draft picks by age

Age Group N Average Age Average GP Outperformance Outperformance (per player) Outperformers
1 84 18.2 53 +687 +8.2 11
2 84 18.4 53 +593 +7.1 8
3 84 18.6 37 +230 +2.7 8
4 84 19.0 44 +297 +3.5 7
5 82 20.1 51 +415 +5.1 10

More evidence in favour of drafting younger players! The first two groups ended up outperforming their (admittedly crude) projection by more than the three older groups did. Each group, on the average, outperformed its projection, partly because in each group there was a player who is now a legit NHL regular (your Michael Ryders and Brooks Laiches) and those players add more than the never-beens subtract. And partly because these players had more junior years to develop, and were not placed immediately into the NHL as the Desjardins equivalencies assume. (Generally speaking, if a fourth-round pick could play in the NHL at age 19, he wouldn’t be on the board in the fourth round.)

Interestingly, just as many of the “really old” players in Group 5 (19.5 or older when drafted) ended up with 30 or more surplus points as the younger groups, which indicates that while the younger players are more likely to outperform in a big way, the older players are just as likely to do better than their projection indicates.

And just as interestingly, the third group sees a huge drop from the second in terms of large outperformance (and Group 4 picks it back up again) and in terms of games played.

But wait a second, what’s that I see in the birthdays of Group 3 players? Oh, look, every last one of them was born in January, February, or March! (I promise didn’t cook the data — it just so happened that way.) Why, could it be that these players, some having already been selected based on their age rather than their ability, are still being selected that way?

Well, it’s not a coincidence. Beware the Q1s.

Conclusions

The two big findings here (well, relative to other findings of mine) are:

  1. Younger players drafted later are more likely to break out than older players.
  2. Players with first-quarter birthdays are still being overrated in the NHL draft.

So, just like in part 1, we see behaviour from hockey executives supporting the hypothesis that younger junior hockey players are undervalued. But this time it’s the NHL teams, not the CHL teams, whose valuation systems are off.

About these ads
This entry was posted in analysis, hockey, junior hockey. Bookmark the permalink.

7 Responses to Junior hockey and age, part 2: the NHL draft

  1. Craig B says:

    Rob, if we look at a population of successful NHL players, say players who play 200 NHL games, or who play 10 playoff games, what is the Qx distribution? Do we know this?

  2. Craig B says:

    Never mind, I see Gabe Desjardins addresses it. I think Gabe misses the point hugely. The bias is still – as these things go – massive.

  3. Craig B says:

    I mean, sure, we won’t need to set up a parallel hockey system like Gladwell reaches for. But clearly there is a substantial bias and it’s worth designing systems to encourage and nurture hockey talent that keep this in mind.

  4. Craig B says:

    Wait, a propos of nothing, did Gladwell *really* write an article on how an incredibly wealthy coach with a group of incredibly rich kids who are all good all-around athletes leveraged the insight and talents of professional athletes and coaches to make them into a very good basketball team? (This was the link I followed from Gabe’s piece). I know it’s easy to pass through the Gladwell Unintentional Irony Filter but wow, that is something. I AM AMAZED, MALC.

  5. Craig B says:

    Holy hell, this Gladwell piece just keeps going and gets more and more unbelievable as it does. Now he’s touting how the press made Charlie Yelverton’s Fordham team into a great basketball team. That’s Charlie Yelverton who went on to play in the NBA.

    Sometimes – no, all the time – I am quite convinced that the only reason I am not even more successful is that I don’t have the unmitigated gall to just pack in even more bullshit.

    • I get e-mail notifications of comments, and there was a “(5)” next to this one in gmail. So imagine my amusement when I kept seeing more and more “Craig B” all down the line.

Comments are closed.