Participation at FIFA age-group events is usually strongly biased towards players born near the beginning of the year and against players born at the end of the year (http://www.bigsoccer.com/forum/showpost.php?p=1446443&postcount=1 for example).
The current olympic tournament has a slightly different pattern. Here are the birth months of the U-23 players in the tournament, broken down into four-month groups:
Jan-Apr: 88 (35.5%)
May-Aug: 104 (41.9%)
Sep-Dec: 56 (22.6%)
Compare this to the breakdown for the 2001 U-20 cup, which is the same group of players:
Jan-Apr: 192 (44.1%)
May-Aug: 155 (35.6%)
Sep-Dec: 88 (20.2%)
So the middle months have caught and supassed the early months between U-20 and U-23, while the final months still lag behind. IMO it's a little surprising that there are still so few late-month players...apparently being born in the middle months is just a short-term handicap, but a late month has longer (permanent?) consequences.
What I found more interesting is the difference between teams that qualified for both tournaments, and teams that just qualified for the U-20.
7 teams from the U-20 tournament also made the Olympics. The birth months for their players in the U-20 tournament breaks down like this:
Jan-Apr: 50 (39.1%)
May-Aug: 50 (39.1%)
Sep-Dec: 28 (21.9%)
17 teams from the U-20 tournament failed to make the Olympics. The birth months for their players in the U-20 tournament breaks down like this:
Jan-Apr: 142 (46.3%)
May-Aug: 105 (34.2%)
Sep-Dec: 60 (19.5%)
The teams that didn't qualify for the Olympics are clearly more biased. An indication that teams that buy success at younger levels by using older players are less likely to succeed down the road?
From the 7 teams that qualified for both tournaments, only 39 players appeared on the roster for both. Some of the turnover is due to overage players taking spots, but it's probably indicative of countries with large talent pools having more success. Players with late birth months were most likely to remain on the team from 2001 to 2004...12 of the 28 (43%) made it to the Olympics, while only 27 of 100 (27%) players born in the first eight months got to Athens.
I've written an article on this subject on the Association of Football Statisticians (http://www.11v11.co.uk) site, looking at English Premiership players. The article relates to Birthdate bias in the Premiership (http://www.11v11.co.uk/page309-233).
The English school year starts in September and there is a heavy bias towards those born early months of the school year in English Premiership players.
One solution would be to divide school sports training into 6 month cohorts.
Is there a reason why Apr-Jun actually seems to do better than Jan-Mar?
The facile answer is that the Jan-Mar players were privileged youths due to their older ages. They were placed on youth teams because of this advantage, they parleyed this youth-team experience to the senior leagues, but then when they got the senior leagues they were overrated. Whereas the Apr-Jun players were good prospects to begin with. (The October - December players were also good prospects, but because they were so young, many of them never made the youth cut in the first place.)
Don't know how anyone would prove that hypothesis, though. Plus, who's to say that this pattern is anything more than one series of numbers in one league?
But thanks for collecting that data, anway. Always good to see actual facts.
One thing you might look at is expanding the cycle from one year to two years. My experience with the American youth players shows a huge frontloading problem with every two year cycle, with the final six months of each two year cycle all but ignored.
For players born mid-60's to mid-70's, the US has a pronounced sawtooth pattern, where even-year productivity spikes up, and odd-year productivity dips back down. I haven't looked at more recent data, but I expect that this will continue to some extent.
In other countries, I wouldn't expect as much asymmetry, since development is less driven by youth national team cycles. That said, Mexico had big crops of talent born in 1973 and 1979, and I don't really know why (chance fluctuation?).
Great stuff, JG.
FWIW, I did some checking sometime around February, and IIRC, MLS did have a shortage of late-month Americans. The Mexican League, however, didn't seem to show any month-to-month patterns.
This might have something to do with the fact that Mexican teams tend to retain a great many youth players, while US soccer thins the ranks pretty quickly.
It could have to do with a temperature difference affecting when Mexican's get it on vs. Americans. A good control would be the same population first. Are American births tilted toward certain months?
NIce work.
The "classic" academic article on this subject can be found in the Sociology of Sport journal, here"
http://www.psychologie.uni-bonn.de/sozial/forsch/ageeffx.htm
You would think the age effect would be pronounced at earlier ages, and then taper off somewhat at older ages, but as this article shows conclusively, the age effect even at professional teams is statistically significant.
I would bet that even those groupings that you show as having a less pronounced age effect in nominal percentage terms are still skewed to older playes in a statistically significant way. Someone should run the set of appropriate statistical tests on the percentages you have presented.
And, of course, the age bias "makes sense." Older stronger more mature kids get picked out as the "best" players at younger ages, they get moved up the food chain faster, get better coaching, more parental committment, head to stronger teams with more visibility and so on, in a relentless self-fulfilling prophecy. While the younger "lesser" player is given short shrift, and gets left behind.
I exaggerate a bit for effect, but not by much.
I believe that all national level coaching staffs worldwide are conscious of this problem, and would like to do something to rectify it, but have no real solution, (assuming that this IS a problem and if so even HAS a solution). In the end, coaches need to win, and you win often with the strongest most mature players -- that is "older" players.
The article that Karl cites shows that age effect (or birthday bias, if you like) is present in professional leagues worldwide.
... present, yes, but to varying degrees and using large samples.
Looking at one Mexican temporada, there was no visible trend, but
I actually have some Mexico data from a longer time window
Mexican players 1995-Clausura 2004
Jan 98
Feb 96
Mar 83
Apr 60
May 74
Jun 65
Jul 79
Aug 73
Sep 73
Oct 69
Nov 52
Dec 61
Non-Mexicans, same time frame
Jan 33
Feb 28
Mar 27
Apr 39
May 35
Jun 32
Jul 26
Aug 38
Sep 31
Oct 27
Nov 24
Dec 37
Among non-Mexican players, I don't observe any trend at all. Among Mexican-born players, it is pretty pronounced. I haven't looked at whether late-year players went on to have more successful careers, though.
Note to swed34: birthrates aren't constant throughout the year, but I'm pretty sure the seasonal variation is relatively small. We would also expect such trends to vary smoothly in time; by contrast, birthdate bias has a spike at the cut-off date.
Wait, what kind of school did you go to? Shouldn't it be that people born in July-September (beginning of the school year) did better than anybody else? I was born in January, but was right in the middle of my class in terms of age.
Depends where and when you are from.
In the old days (i.e., me), the most popular arrangement was calendar year, so as a January birthday you would have been the oldest.
With my son, Sept. 1st is the cut-off date. I think that's a pretty common choice these days.
FYI, a couple of years ago at my son's school they pulled 20 kids out of his grade for advanced math. 16 were born Sept - Dec, only 4 were born Jan - August. The age effect is everywhere. And just as with coaches, very few teachers have a clue about this stuff. Since 1st grade, my son has never had a teacher who was aware of the kids' birthdays.
Another piece of data ... goals scored by Mexican players in the Mexican First Division by month of birth:
Jan-Mar: 1276
Apr-Jun: 1365
Jul-Sep: 1337
Oct-Dec: 935
In terms of goalscoring, the middle month kids overtook the early months, while the late months lag behind. This means that even though more early month kids get the chance to play, they are often less productive.
Goals Scored per Player
Jan-Mar 4.61
Apr-Jun 6.86
Jul-Sep 5.94
Oct-Dec 5.14
Didn't have time to finish the final column before ...
Mons Num Min Min/Plyr
Jan-Mar 277 1,317,453 4,760
Apr-Jun 199 1,143,632 5,750
Jul-Sep 225 1,160,492 5,160
Oct-Dec 182 854,228 4,690
Is there a reason why Apr-Jun actually seems to do better than Jan-Mar?
Although a little off topic, I thought I might add another log to the flame...
I did an unscientific poll years ago in college with a bunch of friends. We made a list, maybe 50 deep, of friends we all had from HS whose birthdays we knew. We then wrote down where they went to college.
What we concluded was people born earlier in the year (jan-mar) attended or were accepted to schools with better reputations. (at least had a lower acceptance rate) ie. Ivy league, Duke, Stanford, Notre Dame, etc.
Apr-Jun birthdays got into slightly less noteworthy, but still strong programs. The trend continued where students born between Oct-Dec (youngest in their respective class) performed lower on SAT scores and on average attended less demanding colleges. It was my contention that, on average, the youngest people in a group are at a significant disadvantage. And in our little sample back then it proved itself out.
I always felt disadvantages exsisted for "born late in the year" soccer players as well, but never had any hard proof. This thread could have been of great use to me years ago to settle a few more arguments. :D
Wait, what kind of school did you go to? Shouldn't it be that people born in July-September (beginning of the school year) did better than anybody else? I was born in January, but was right in the middle of my class in terms of age.
One thing you might look at is expanding the cycle from one year to two years. My experience with the American youth players shows a huge frontloading problem with every two year cycle, with the final six months of each two year cycle all but ignored.
Good work, though.
In terms of minutes played, we don't see the phenomenon of middle month players overtaking early month players ...
Mons Num Min
Jan-Mar 277 1317453
Apr-Jun 199 1143632
Jul-Sep 225 1160492
Oct-Dec 182 854228
Although a little off topic, I thought I might add another log to the flame...
I did an unscientific poll years ago in college with a bunch of friends. We made a list, maybe 50 deep, of friends we all had from HS whose birthdays we knew. We then wrote down where they went to college.
What we concluded was people born earlier in the year (jan-mar) attended or were accepted to schools with better reputations. (at least had a lower acceptance rate) ie. Ivy league, Duke, Stanford, Notre Dame, etc.
Apr-Jun birthdays got into slightly less noteworthy, but still strong programs. The trend continued where students born between Oct-Dec (youngest in their respective class) performed lower on SAT scores and on average attended less demanding colleges. It was my contention that, on average, the youngest people in a group are at a significant disadvantage. And in our little sample back then it proved itself out.
I always felt disadvantages exsisted for "born late in the year" soccer players as well, but never had any hard proof. This thread could have been of great use to me years ago to settle a few more arguments. :D
Although a little off topic, I thought I might add another log to the flame...
I did an unscientific poll years ago in college with a bunch of friends. We made a list, maybe 50 deep, of friends we all had from HS whose birthdays we knew. We then wrote down where they went to college.
What we concluded was people born earlier in the year (jan-mar) attended or were accepted to schools with better reputations. (at least had a lower acceptance rate) ie. Ivy league, Duke, Stanford, Notre Dame, etc.
Apr-Jun birthdays got into slightly less noteworthy, but still strong programs. The trend continued where students born between Oct-Dec (youngest in their respective class) performed lower on SAT scores and on average attended less demanding colleges. It was my contention that, on average, the youngest people in a group are at a significant disadvantage. And in our little sample back then it proved itself out.
I always felt disadvantages exsisted for "born late in the year" soccer players as well, but never had any hard proof. This thread could have been of great use to me years ago to settle a few more arguments.
So, just out of curiosity, I decided to look at the academic trend from a single-college perspective. The first thing I thought of was my class directory, which would have given the birthdays of 200+ people who got into one of America's most demanding schools (and no one's going to argue that Caltech isn't one)... but I can't find it right now. I had to settle for my current Facebook list, which is fortunately pretty extensive! I took the birth months of everyone at Caltech who listed birthdays.
Jan 3
Feb 8
Mar 4
Apr 2
May 1
Jun 4
Jul 6
Aug 4
Sept 6
Oct 3
Nov 4
Dec 6
Breaking into quarters using Sept. 1 as a cutoff date (it's the norm for American schools today - my sister, born in late July, was one of the youngest in all of her classes), we find...
Sept-Nov 13 (11)
Dec-Feb 17 (13)
Mar-May 7 (7)
Jun-Aug 14 (13)
The numbers in parentheses are after removing both those who graduated from high school at least a year early and those who were held back a year at the beginning. Their birth months are:
Graduated early (5): Feb, Feb, Jul, Oct, Dec
Held back (2): Sept, Dec
Either with or without those seven, the youngest are easily holding their own, but the third quarter is underrepresented. Haven't checked to see if statistically so, though. Can anyone come up with a plausible explanation other than seasonal birth rate variations?
Among non-Mexican players, I don't observe any trend at all.
Odd. I have no explanation. Perhaps the non-Mexican players come from countries with various start/end dates for their youth soccer systems? If so, then the age effect would be masked.
Note to swed34: birthrates aren't constant throughout the year, but I'm pretty sure the seasonal variation is relatively small.
Well, it's bigger than I would have thought, as births seem to fluctuate by as much as +/- 10% in the U.S. from month to month. But the birthdate bias is much, much larger than +/- 10%.
One solution would be to divide school sports training into 6 month cohorts.
Way OT....
But I think this would be a good principle in general. I had a niece who stayed with us for a year with an August birthday, and was practically the youngest kid in her class. She really struggled, and that's only natural...the oldest kids in a kindergarten class are, what, 16% older? I have a daughter whose birthday comes 8 days after the deadline here in NC. Of course, alot of parents with kids born just before the deadline will hold their kids back, so my girl won't be quite the oldest kid with a huge, fake advantage her first years in school. But really, they ought to have schools run on half-year tracks. Maybe they could close the gap in, say, the summer between 5th and 6th grade, by having the "behind" kids go to summer school and catch up. But as a parent, I'm almost tempted to push my girl, if she's as smart and advanced as her brother, to start school early.
FWIW, I did some checking sometime around February, and IIRC, MLS did have a shortage of late-month Americans. The Mexican League, however, didn't seem to show any month-to-month patterns.
This might have something to do with the fact that Mexican teams tend to retain a great many youth players, while US soccer thins the ranks pretty quickly.
The article that Karl cites shows that age effect (or birthday bias, if you like) is present in professional leagues worldwide.
If Mexico is an exception, then Mexico is unusual indeed and should be analyzed for its differences vs. the rest of the world, not just the U.S.
FYI, the U.S. ODP program is now consciously adjusting for the age effect, at least at the younger ages. For example, in Illinois the '91s, '92s, and '93s are evaluated initially by their performance in 6-month age brackets.
At those ages, 6 months tends to make a huge difference. For example, the a late '93 Illinois team that was assembled from the kids who showed up at a summer tryout lost to an early '93 team by a score of 7-0.
In other countries, I wouldn't expect as much asymmetry, since development is less driven by youth national team cycles.
Here's the percentage of odd-year players at the last WC, broken down by confederation:
AFC 59%
CONCACAF 58%
CAF 50%
CONMEBOL 50%
UEFA 48%
Africa's percentage is somewhat skewed because Nigeria had the most even-year players in the tournament, 18 out of 23--take them out and the percentage of odd-year players jumps to 58%. Taking that into account, the "odd-year effect" is more pronounced in the weaker confederations, where the clubs have fewer resources and we might expect youth NT play to be more important for developing players.
IMO it's a little surprising that there are still so few late-month players...apparently being born in the middle months is just a short-term handicap, but a late month has longer (permanent?) consequences.
Great stuff, JG.
FWIW, I did some checking sometime around February, and IIRC, MLS did have a shortage of late-month Americans. The Mexican League, however, didn't seem to show any month-to-month patterns.
This might have something to do with the fact that Mexican teams tend to retain a great many youth players, while US soccer thins the ranks pretty quickly.
 Walking/Running Puppy - Advise needed
 GridPanel border
|
Birthdate bias and youth team success: from U-20 to U-23