May 12, 2006

Levitt: "Maybe the World Cup Wasn’t the Best Example"

On their Freakonomics blog, Levitt and Dubner step back from their claim in the New York Times that World Cup soccer players are mostly born in the first six months of the years. They offer some persuasive data that NHL hockey players, however, tend to be born in the first half of the year by a 59-41 ratio. A good discussion ensues.

A reader sent me the following showing a 58-42 bias toward first half of the year births in the NHL's top ten draft picks.

He wrote:

You had asked if anyone had any data on Hockey players. Well, I did some non-scientific research of Birthdays of the Top Ten drafted players from each of the last 10 drafts. And this is what I found:

Jan 11
Feb 14
Mar 8
Apr 14
May 6
Jun 5
Jul 3
Aug 7
Sep 14
Oct 7
Nov 7
Dec 4
Sum 100

A few pieces of trivia about Hockey players that might have some affect on who gets drafted.

- Few players get drafted after the age of 18. A few players get overlooked and then shine in college or elsewhere, but this is not often.

- You can not be drafted before the age of 18.

- Three leagues/tournaments play a huge role in who gets the most attention:

* The Canadian Major Juniors (OHL, WHL and QMJHL)

* The Under-18 Tournament (U-18)

* The World Junior Championships (WJC)

If you do not make a major showing at one of these places, you will not be a top draft pick. The only exception would be a European/Soviet-Bloc player who plays in one of their Elite leagues but was injured for the WJC.

I should also note that I did not factor for NHL success. That is, there are plenty of players who are drafted in the top ten who never make it in the NHL because they never mature/improve. With some more time this could be done though. Simply go to HockeyDB.com and view each draft, the site has their career stats with total number of games played. This is not that helpful for recent picks, but would be helpful for player drafted in the late 80’s early 90’s.

Anyway, here is where I got my data: http://www.hockeydb.com/ihdb/draft/index.html (warning: lots of pop-ups)

So, in hockey, both in Canada and abroad, there appears to be a severe bottleneck to advancement at roughly ages 17-18, which gives a big advantage to boys who were born in January and thus are 11 months older than boys born in December.

We see a lot of the same thing in horseracing, where the Kentucky Derby is restricted to 3-year-olds. I recall watching a fictional TV show as a kid about horsebreeders with a mare going into labor on New Year's Even, so they trucked her across the time zone line so she would give birth in the right year so that her foal would be 364 days more mature when he was eligible for the Kentucky Derby.

So, the question I raised on Levitt's blog is: Is this bias in the NHL a self-esteem effect as Levitt theorized, or a market inefficiency?

David Kane posts on Freakonomics:

The more that I look at this, the more it seems clear that, while Dr. Levitt is correct that more NHL players are born in the first half of the year, this has nothing to do with the stars-are-made-not-born thesis of the article. In other words, the reason that there are more early-births is not that early-births get more practice against better opponents. Instead, early-births are more likely to get high profile spots which give them exposure for the NHL draft. They are more likely to be drafted than a similarly talented late-birth because the coaches and GMs have seen them play as (more highly developed) teenagers.

In other words, the hockey draft is inefficient. Teams should draft fewer early-births and more late-births.

How might we test this? Easy. The better the cohort of players, the less inefficient the market will be. Among the best players (those who get lots of ice time for several years in the NHL), there will be no meaningful difference in early-versus-late births. These players are judged accurately on their adult skills.

Instead, the effect will be much greater in the bottom of the NHL pool. Younger players with not a lot of ice time are more likely to be judged and retained on the basis of their (inaccurately measured because of birth-month issues) performance in youth leagues.

If I am correct, the effect will be smaller and/or non-existent among older, better players. Alas, I can’t get any of these data sources to provide a clean test of this hypothesis, but I was able to split the Sportsnet.ca data into two portions: Jan-Jun and Jul-Dec birthdates. When you do this, the default sort is by points scored.

I picked 30 points as a reasonable threshold. Turns out that, among the 174 NHL players who have scored at least 30 points this season, 140 were born in Jan-Jul and 134 in Jul-Dec. (Of course, points scored is not the best measure. What about goalies and defensemen? And so on.)

But, big picture, there is no birth-month effect among the top 1/3 of NHL players. This suggests to me that the birth-month effect is much more likely to be a draft inefficiency. You don’t see this in other sports because the draft process is better.

A very interesting analysis. We know now that there were a number of market inefficiencies in baseball, as identified by Bill James from 1975 onward. So, this is at least theoretically possible. But I don't know anything about hockey statistics. I wouldn't say Kane's tabulation is terribly definitive (after all, there is a small effect in the direction of Levitt's theory), but this is a good example of what blogging can accomplish.

UPDATE: Kane adds:

Using the Sportsnet.ca site, it seems that the birth-month effect is just as strong among older players as it is among young players. For example, of the 284 players who are 30 years old or older, 161 were born in the first half of the year and 123 were born in the second.

Hmmm. I would have guessed that older players would be evaluated more efficiently than younger players and so we should not see a birth-month effect. Then again, a different way of looking at the problem is to note that all 30+ year olds had to go through the same inefficient draft.

As I said, I don't know how this will shake out empirically. I suspect there is a bottleneck effect that varies by sport and by position. For example, if you have potential to make it to the major leagues as a baseball outfielder, you are almost certain to get a lot of playing time in high school, since very, very few high schools have four outfielders with major league potential .

On the other hand, only one quarterback can start at a time. The most famous quarterback bottleneck of all time was on the San Francisco 49ers from 1987-1992, when they had both Joe Montana and Steve Young, who just might be the two best quarterbacks ever. If Montana had been more durable, who knows if Young would ever have gotten a chance to show what he was capable of?

So, there's a definite "opportunity effect." For example, the younger brother of a friend of mine was a standout high school quarterback, and got a scholarship to Stanford. But Stanford had a quarterback named John Elway, so he sat on the bench for three years. Finally, Elway graduated, so he became the starter for Stanford his senior year ... and didn't deliver. So, the NFL didn't show any interest. He went into minor league baseball when he got drafted in the 14th round by the SF Giants. Maybe he never really was that good at quarterback. But maybe he just needed a year to stink up the joint and learn from his mistakes, so if he'd gone to, say, a Utah St. and been able to start as a sophomore, he would have been pretty good as a senior.

If you want to be an NFL quarterback, it can seem like a good idea to attend a quarterback factory high school like mighty Hart H.S. in Southern California. (Here's an extremely objective Wikipedia article on Hart grad and NFL quarterback Kyle Boller.)

But, that’s a more severe bottleneck because you might very well end up sitting on the bench throughout your entire high school career because another NFL potential quarterback is starting ahead of you. And if you don't start in high school, it's hard to get a scholarship to a football factory college, which makes it really hard to get to the NFL.

Thus, it can be very helpful if your parents redshirt you as a child, making you spend two years in preschool, so you'll then be a year older than your classmates for your entire childhood. For example, this year's most hyped high school junior QB, Jimmy Clausen of Oaks Christian in Westlake, CA, who just signed with Notre Dame, "started kindergarten at six and repeated sixth grade, "to gain maturity," says his mom, Cathy."

This raises the related issue of whether you are best off having less competition or more when growing up. For example, Jimmy Clausen is both older than other high school quarterbacks, and his parents sent him to a small school that plays only other small high schools. So, comes out and throws four or five touchdowns in the first half, then sits out the second half. This is great for his self-esteem, but is it toughening him up enough? When he's at Notre Dame and playing USC and with three minutes left in the game he's completed only 11 of 29 for 122 yards with three interceptions and he's got to drive the Fighting Irish 80 yards for a touchdown, is all that self-esteem in high school going to do him a lot of good? Well, ND coach Charlie Weis seems to be more focused on Clausen's Dan Marino-like release than such issues. So, talent matters.

Here's my 2002 article "Redshirting: A Kindergarten Arms Race" on the trend toward American upper middle class parents trying to give their kids a leg up on the competition by holding them back for a year.


My published articles are archived at iSteve.com -- Steve Sailer

No comments: