Are stars born or made? Here’s what World Cup 2006 has to say on the issue.

Last month we wrote a New York Times column asking whether superstars are born or whether they are made through a combination of the lots of practice, the right kind of practice, and coaching. The experts in the area suggest that superstars are made.

One conjecture we made in that column was that because the FIFA cutoff date for determining a child’s age for international play is January 1, we would expect that a disproportionate number of the players in the World Cup would be born in the early part of the year. The idea is that these kids will get special treatment and attention when they are teenagers because they will be developmentally almost a full year ahead of kids born in the latter part of the year. These developmental differences will fade away with age (i.e. by the time you are 25 it doesn’t matter much if you are 25 years and 1 month old or 24 years and two months old), but the early success, access to coaching, playing experience that the older kids got as teens would have long term effects on their soccer careers.

This conjecture led to some vigorous blog commenting (see here and here).

There were two nuances we missed though. First, the January 1 eligibility date wasn’t put into place by FIFA until 1997. And also, individual countries use different month cutoffs for determining who is eligible for different leagues within the country. Many countries use a date in the fall for their cutoffs, but there are a whole range of eligibility dates.

So, the clearest prediction we are left with concerning the World Cup is that if the FIFA January 1 cutoff matters and the experts in this area are correct, then for players born late enough to be affected by the 1997 date change should be disproportionately born in the early months of the year, and much less likely to be born in the last few months of the year. For players who are too old to be affected by the 1997 FIFA rule change (i.e. they were not teenagers at the time of the change), it is unclear what the pattern will be, but there is certainly no reason to expect a lot of January players. To really do these older players right, one would have to go country by country to determine what the cutoff rules were. (Something we haven’t done, but which I suppose some World Cup crazed blog readers will decide to do.)

We found a spreadsheet with the dates of birth of all players on World Cup rosters in 2006. What do the data say?

First, among players born in 1979 or later, making them 18 or younger when the FIFA January 1 cutoff took effect, the months of birth look like this:

January — 52
February — 43
March — 35
April — 38
May — 38
June — 25
July — 29
August — 31
September — 26
October — 35
November — 22
December –27

32.4 percent of the players were born in the first three months of the year, 25.2 percent of the players were born in months four through six, 21.5 percent in july to september, 21 percent in the last three months. Exactly what the theory predicts.

How about for the older World Cup players, those born before 1979? A very different pattern emerges:

January — 22
February — 23
March — 22
April — 23
May — 27
June — 31
July — 29
August — 36
September — 34
October — 27
November — 32
December –29

For the older players, only 20 percent of the older players were born January to March. 24 percent were born April-June. The months July-September were the most common.

All in all, this seems like pretty strong evidence in favor of the theory.


Leviathan

1997 1979?

Harlan

Interesting! That's a pretty clear pattern. It reminds me of another phenomenon, which is handedness in several sports. In fencing, which I used to compete in, a much-greater-than-chance proportion of international-level fencers are left-handers. Now, anyone competing at that level is fully able to deal with left-handed fencers, and there's not much reason to think that there's any real advantage. (Unlike, say, baseball, where southpaw pitchers can throw curves that go the other way.) However, at the beginner levels, a left-handed fencer is a novelty to other fencers, and so the left-hander has a big early advantage in results. They may not be any better than their right-handed peers, but the handedness throws people off and they so get extra reinforcement early on and tend to stay in the sport longer. At least, that's the theory.

mjbigelow

In the data, it's clear that Jan and Feb are clear leaders but April and May both have more than Mar. It seems to dilute the effect to include those two months.

Also, among the older generation (not born in 1979 or later), the opposite pattern emerges, with June through December having the largest counts. Could it be that because those born later in the year are somewhat younger than the rest of their cohort, they stay in the league longer? It seems like that would only affect somebody mid-year, not over time.

Or is it just a coincidence?

TartanBill

Don't dwell on individual months, the sample size is too small.

Using all the data, a very simple calculation from the post 79 group yields
y = -1.89x + 45.7 with R^2 = 0.63

where X is the number of the month born in the year and Y is the number of players in that bin. That's a pretty steep slope and a fairly high correlation. Although January looks like an outlier, removing it only modestly weakens the trend.

For pre 79, the same analysis yields
y = 0.96x + 21.7 with R^2 = 0.52. The slope is not as steep and the correlation more modest.

In both cases, the trend is significant enough to lead one to believe there is an underlying cause. Of course alternative explanations exist. It may be that stars are unmade when the later birth month children lose interest and quit early.

beach

Harlan,
Is that tendency true across all weapons? I used to fence, too-- at the collegiate , not international, level. Even in my days as a newby, I never had trouble with left-handers thanks to my coach. He always said, "Hit lefties on the face mask." It worked every time, and I even beat the Spanish National Champion once using that little trick. However, only epees like me could use that strategy, and I always wondered how new fencers in the other weapons dealt with lefties. Perhaps they didn't?

sppeterson

Is this already accounting for general tendencies for more people to be born in certain months?

I had heard somewhere that births were more common in the August to October range since couples see all these families enjoying Christmas and decide to start a family of their own.

SteveSailer

I would guess that all World Cup players, no matter what advantages they received in nurture, are at least two standard deviations above the mean in natural ability compared to their compatriots. So the nurture glass here is 2.3% full and 97.7% empty.

will

Thanks for the guess SteveSailer. Now people can stop doing research and just ask you what your guess is in the future.

Or did you have some evidence to support your "guess". I guess not.

John S.

This seems like a perfect situation for a Chi-square test, is anybody going to do one?

dkane

You originally wrote:

----------------
If you were to examine the birth certificates of every soccer player in next month's World Cup tournament, you would most likely find a noteworthy quirk: elite soccer players are more likely to have been born in the earlier months of the year than in the later months.
----------------

This is false, as I was the first to demonstrate. Now, the stand-up thing to do, having been proved wrong, is to, first, admit that you were wrong. Fess up. Don't try to hide the fact that your prediction was not born out by the facts.

As a side note, if you saw my comment before posting this analysis, the decent thing to do would be to credit me. (I realize that you may not have; perhaps, you don't read all the comments on your own blog.) But, since you have still failed to give public credit to the astrology site from which you stole (without attribution!) your hockey graphic, I am suspicious of the standards that you use to credit the work of others. Did you see my comment?

But, back to the issue at hand, having been proved wrong, you now spend a bunch of time looking at the data, playing with the data, trying to come up with some breakdown that supports your claim. Eureka! You found one! Was this the first analysis that you tried? Was it the only one? This reeks of the worst sort of datamining. Unless you provide (rough) details of the tests that you did on this data, there is no way to evaluate the statistical significance of the single result that you report above.

Perhaps Bill Lloyd will turn a careful eye toward some of the claims that you make here.

Without looking at the data, I have a guess to make. We all agree that, for younger players, being born early in the year is an advantage. So, the stars of the under-20 team are more likely to be early-months.

Now, when picking non-starting players to fill out his roster, every World Cup coach should think about future international competitions. You want to bring along your best future players so that they gain experience for the next time around. So, there will be a significant birth-month effect among younger non-starting players on World Cup teams because coaches do not do a good job of adjusting their forecasts of future stardom for age. They just pick the best under-20 (or whatever) players without paying attention to the soon-to-disappear advantage of an early-birth.

This is analogous to claims (made by Steve Sailer and others) that the NHL birth-month effect is driven by an inefficiency in the NHL teenage draft.

Prediction: Among World Cup starters, there will be no birth-month effect, whether for older or younger players.

In a few days, we will have enough data to test this. Stay tuned.

Read more...

JanneM

Two things hit me: first, (as another poster already commented), what is the normal birth frequencey pattern? Births are not evenly distributed over the year after all.

Second: when is the season for football? My guess is that at least in countries with distinct seasons, your birthdate in relation to when the season starts will have more effect than the FIFA rule. Indeed, you may see a stronger effect than expected in places like Scandinavia, with those who are born at the "wrong" date for the start of football season opting out of the sport altogether in favour of ice hockey or skiing (for which they'd be calendarically much better suited).

okansas

A similar analysis for baseball (my appologies if it has already been linked):

http://www.sabernomics.com/sabernomics/index.php/2006/05/age-cut-offs-and-month-of-birth-in-baseball/

I thought Anders Ericsson's work on deliberate practice was the more interesting part of the NY Times article. I came across an article about an machine that measures the angle of a basketball free throw and then reports it immediately to the player practicing. It sounds like a way to do some deliberate practice (and doesn't sound like it would be much fun):

http://www2.kusports.com/news/2006/jun/13/rush_scores_passing_grade/?mens_basketball

lllach

Beware: as explained in lacienciamaldita.blogspot.com, countries where the share of post-1979 players born in the first quarter is over 50% are Ukraine, Brazil, Togo and Tunisia. Famously, some soccer federations alter their players' birth dates for youth championships. Keep in mind that the whole anomaly is that out of 400 post-1979 players, 130 were born in Jan-Mar, rather than the statistical average of 100/quarter. We need just 10/20 cases of rigging with FIFA regulations for the anomaly to become statistically insignificant.

John S.

I finally found statistics on birth months. For the United States in 2000, they are here, in Table 15:

http://www.cdc.gov/nchs/data/nvsr/nvsr50/nvsr50_05.pdf

The distribution is:

January 8.13%
February 7.82%
March 8.39%
April 7.81%
May 8.41%
June 8.41%
July 8.60%
August 8.87%
September 8.56%
October 8.47%
November 8.22%
December 8.30%

This allowed me to do a chi-square test on the players' birthdates. For those born after 1979, the p-value is 0.002742. In other words, it is very unlikely that the distribution of these players' birth months is the same as the U.S. 2000 birth month distribution.

On the other hand, for the players born before 1979, the p-value is 0.8308. There is no evidence to suggest that these players' birth month distribution is any different from the US 2000 distribution.

Leviathan

1997 1979?

Harlan

Interesting! That's a pretty clear pattern. It reminds me of another phenomenon, which is handedness in several sports. In fencing, which I used to compete in, a much-greater-than-chance proportion of international-level fencers are left-handers. Now, anyone competing at that level is fully able to deal with left-handed fencers, and there's not much reason to think that there's any real advantage. (Unlike, say, baseball, where southpaw pitchers can throw curves that go the other way.) However, at the beginner levels, a left-handed fencer is a novelty to other fencers, and so the left-hander has a big early advantage in results. They may not be any better than their right-handed peers, but the handedness throws people off and they so get extra reinforcement early on and tend to stay in the sport longer. At least, that's the theory.

mjbigelow

In the data, it's clear that Jan and Feb are clear leaders but April and May both have more than Mar. It seems to dilute the effect to include those two months.

Also, among the older generation (not born in 1979 or later), the opposite pattern emerges, with June through December having the largest counts. Could it be that because those born later in the year are somewhat younger than the rest of their cohort, they stay in the league longer? It seems like that would only affect somebody mid-year, not over time.

Or is it just a coincidence?

TartanBill

Don't dwell on individual months, the sample size is too small.

Using all the data, a very simple calculation from the post 79 group yields
y = -1.89x + 45.7 with R^2 = 0.63

where X is the number of the month born in the year and Y is the number of players in that bin. That's a pretty steep slope and a fairly high correlation. Although January looks like an outlier, removing it only modestly weakens the trend.

For pre 79, the same analysis yields
y = 0.96x + 21.7 with R^2 = 0.52. The slope is not as steep and the correlation more modest.

In both cases, the trend is significant enough to lead one to believe there is an underlying cause. Of course alternative explanations exist. It may be that stars are unmade when the later birth month children lose interest and quit early.

beach

Harlan,
Is that tendency true across all weapons? I used to fence, too-- at the collegiate , not international, level. Even in my days as a newby, I never had trouble with left-handers thanks to my coach. He always said, "Hit lefties on the face mask." It worked every time, and I even beat the Spanish National Champion once using that little trick. However, only epees like me could use that strategy, and I always wondered how new fencers in the other weapons dealt with lefties. Perhaps they didn't?

sppeterson

Is this already accounting for general tendencies for more people to be born in certain months?

I had heard somewhere that births were more common in the August to October range since couples see all these families enjoying Christmas and decide to start a family of their own.