Age Bias in Brief

Age bias in drafting is a very hard thing to describe for two main reasons: there's two separate biases acting against each other, and the reality goes against common wisdom, so is therefore most likely to be discounted.

In trying to explain escrow when it becomes necessary to talk about – which is going to happen soon when a new CBA is done – I went with the simplest point form there is. This leaves out all the background, details, examples, implications, etc. This is going to be something similar for age bias in the NHL, a more appropriate title because it affects the choice of who plays and how much in the NHL well beyond the draft.

Background Facts

1. The NHL draft cut-off date is September 15.
2. The birth month of draftees in their first year of eligibility, oldest to youngest is: Late September, October, November, December, January, February, March, April, May, June, July, August, Early September.
3. Most junior hockey programs use a calendar year age system.
4. The two countries that produce the vast majority of NHL players, Canada and the United States, both show births peaking in the summer with February the month with the least births (fewest days) and winter months in general filling out the lowest counts. The differences are not large.

The Wrong Assumption

Draft picks are discussed constantly, and mostly wrongly, by birth year. This year, we are seeing 2006 and 2007 birth years. The 2006 people are the ones with late-September to December birth months. They are the oldest players in the draft (at all times in this discussion, I'm leaving out overagers and re-entry players). The 2007 birth years cover almost three-quarters of the year, and range from nearly as old as the 2006 players to the very youngest in the draft.

The long-standing common wisdom is that the oldest players with the earlier birth year are overrated by their results because they are "beating up on younger players" in junior hockey. They should be skitched down in the draft rankings to account for this.

This is completely wrong both in the assumptions about how relative age affects results and in how age bias in general and in hockey specifically actually manifests itself.

What is Meant By Bias Here?

The word bias gets confused with hate or bigotry because of its use in ordinary conversation. What bias means here is a belief or assumption that is factually wrong or a result that is skewed from the expected outcome that is due to that belief that might not be conscious. The bias can arise from people acting in reasonable, if self-interested ways, or it can be irrational – that is arising out of actions that go against self-interest.

What are the Expectations?

Given the distribution of births, and therefore people available to choose to play hockey, be dedicated and skilled, end up in elite leagues as children, get noticed by scouts and end up in the NHL someday, we should expect most of the NHLers to be born in the summer, with fewer from the winter months.

What is the Reality?

This is easy to answer for the NHL by team, by season or by decade if you like. Both Quant Hockey and Elite Prospects will give you instant charts on demographic information. This is the current regular season in the NHL:

Quant Hockey's totals page which contains thousands of players over many seasons shows the same general result. January produces more NHLers than any month, and overall, February, which the fewest people born, is third.

On draft day a player born on December 31 is one day older than the one with a birthday on January 1. And yet this season, there were almost twice the January babies as December.

This is a bias.

The Older Players are Overrated Cliché Can't be the Cause

The simple answer that scouts and teams don't draft as many September 15 to December 31 players because they think they're overrated on draft day can't explain the chart above. There's a host of reasons why this can't be the smoking gun – the phenomenon affects the summer birthdays by a greater amount since they have the highest birth rates. If it was the draft cut-off at fault, explain why August is so low, when it should be higher as the youngest underrated players reside there according to this common wisdom.

But much more meaningful than those caveats on this simple and wrong explanation is this, the OHL this year:

Which is broadly representative of all feeder leagues into the NHL, including those in Europe.

Summary So Far

• Junior leagues are grossly overrepresented by players born earlier in the year declining to very few players born late in the year.
• This is not how population distributions work, they peak in the summer.
• This distribution persists into the NHL, but is usually less extreme.
• The draft cut-off date ideas about who has juiced stats and who is behind in development of draft day may make this worse, but can't explain the underlying bias.

It's Not Just Hockey

There is a great deal of evidence, primarily from education, that birth month and therefore relative age in groups of people influences how they are assessed. Teachers tend to favour the children with earlier birth dates and disfavour the late birthdays. There are many theories about this involving relative physical and mental and emotional maturity, but this is a widely understood and widely experienced phenomenon.

Rational or Irrational

I think it's pretty obvious that an OHL team filling out its roster may well choose at their draft the older, bigger, more mature players. That's a rational self-interested reason for the OHL age bias.

But long before they draft players, the age bias has been funnelling players onto better lines, better teams, better leagues in a biased way.

Birth month distribution in the draft is just as stark as in the feeder leagues but on NHL draft day, the oldest players are from the late months of the year. If the bias were rational-seeming like it is in the feeder leagues, those months would see more players taken.

Real Research Accounts for Outcomes vs Draft Position

A simple look at the distribution of birth months in the NHL just hints that there is a bias. Deep research into this issue shows that the later a player's birthday is in the calendar year, the more likely they are to outperform their draft ranking, but conversely, the less likely they are to even be available at 18 to be drafted.

In other words, the bias has removed potential people from the sport as children, kept them out of the leagues and teams they should be in and they end up being drafted later, and therefore getting fewer opportunities in the NHL. The higher you're drafted, and this is another bias well studied, the more chances you get to play regardless of your actual output.

Biases Stack

There is some evidence on the research into age bias that the NHL draft actually manages to reverse some of the effects and does better at drafting than they perhaps should given the biased inputs. The age bias still exists, and the wrong-headed business about birth year will persist forever, but the NHL actively quantifies prospect skills in ways to fight against this bias.

However, age bias stacks with size bias, and might even be the same thing for younger children playing the game. And, of course, the pool of available players comes to the NHL draft heavily biased.

Your Mental Correction is Wrong

You cannot, alas, take understanding of this bias and accurately apply some corrective to individuals. This sort of bias can be seen in large groups – like the 1,000 or so NHL players in a given season or the 200+ players drafted each year. You can find it in the last 10 years of top-15 drafted players. But what you can't do is look at a couple of guys in the draft and rearrange them just based on birth date.

Bob McKenzie's mid-season draft list had Matthew Scheafer (September 5) first and James Hagens (November 3) second.

By the conventional (and wrong) draft age thinking, Hagens should be ranked too high because he's a lot older than most players, and it makes him look better. By the actual way age bias works throughout a player's life, Hagens should be ranked too low because he was born later in the year.

In the final McKenzie rankings Hagens fell to seventh. And while we can't discount that some of the thinking may have changed based on the idea he was older that some others and maybe not as good as he appeared, we can't assume that's the case. Hagens likely did have to push against some bias in his short career to date, that doesn't make him better than Shaefer. You can't apply a simple corrective because you will, inevitably overcorrect.

How Can you Utilize This Understanding?

This is key. Research for the sake of it is all well and good, but you need to be able to use information for it to be valuable.

I think the best way to overcome the bias in a scouting department would be to outright ban the discussion of birth year and to make sure age isn't being used to weight draft analytical systems incorrectly.

Once that's done, the smart move is to start looking for players born late in the year playing in the lesser NCAA or CHL teams, lower in lineups, getting less buzz. The kind of player there's stories about to explain their good results and lack of opportunities, stories that might just be justifications. Apply some less subjective scouting analysis to those guys and see what you can find.

After all, that 15% of the NHL that are undrafted players has to come from somewhere.