The Matthew Effect: Talent ID and sports science application

The Matthew Effect: Talent ID and sports science application

The first post of 2009 is inspired by a book I read over the break - Outliers by Malcolm Gladwell. One of the very first things I recall learning when I started out my postgraduate studies was a tip from Prof Tim Noakes to read widely, and read outside your field. Sensible advice, and it informs much of what we are about here at The Science of Sport.

And in that light, Malcolm Gladwell's book, Outliers, provides the latest inspiration for a somewhat novel look at sports science and talent ID (it's "somewhat novel" because it's actually a well-established phenomenon, but has not been published in sports science journals, but mainly social science journals). Gladwell is not entirely outside my field, it must be said, because much of it deals with economics, marketing, strategy etc. But, nevertheless, it's a different perspective on sports science, especially when it covers such obvious territory as the topic of today's post.

That territory, for those of you who have read Outliers, is called The Matthew Effect.

The Matthew Effect - lessons from ice-hockey, soccer and rugby

Today's post is really just a summary of what is described in the book, with the addition of an example from South African rugby that I've managed to do so far, and my own interpretations. There is much more to come, however, including more examples (some of which I hope you'll provide), as well as discussion of how different sports might be affected by this phenomenon, what might be done, and what the implications are for sports science and management.

But for now, forgive me for merely summarizing the data provided in the book, and take a look at the following set of pie-charts. What you are looking at are the breakdown of month of birth of junior players in four different high level sports teams. From top left, going clock-wise: A Canadian Junior Champion team from 2007, the Czech Junior Football/Soccer team from 2007, the Czech Junior Ice-Hockey team from 2007, and the 2007 South African Schools Rugby team.

What should jump out is the enormously high percentage of high-level players who are born between January and April. All told, out of the 91 junior players making up the above four teams, 55 of them (60%) were born in the first four months of the year, and only 13% in the last four months.

This is not an isolated finding, and is true across just about any sport at a high-level. It was reportedly first observed in the mid-1980's by a Canadian psychologist named Roger Barnsley. He noticed that a disproportionately high percentage of high-level ice-hockey players were born in the first few months of the year, and almost none towards the end of it. He expanded his study and looked at other sports like football and baseball, and even started to examine the effect of birth-month on things like academic achievement, suicide and self-esteem. You can read some of those studies here.

The reason - relative age, and a confusion of ability with maturity

We know you're pretty sharp, so it will probably come as no great surprise to learn that this finding is the result of the effect of RELATIVE age on sports performance, and the very easy mistake that coaches are lured into making. If you thought that it was the result of their star signs and some astrology, then I'm afraid you were on the wrong track! But, you might enjoy this website a little more...

Back to reality, and the suggestion that the reason so many elite athletes are born in the early months of the year is the result of the very large effect that 10 months difference in age can have on young children's ability to play sport.

An example: Ross and Jonathan

Let's take two 10-year olds, Ross and Jonathan. They are both 10 years old on the first day of January 2009, and so they compete in the Under-11 age group of their sport (soccer, let's say).

However, not all 10-year olds are created equal! Ross is 10 years and 11 months old on the first day of 2009 (his birthday is in February). Jonathan is 10 years and 1 month old (his birthday is in December), and so he is a full 10 months YOUNGER than the people, like Ross, who he is going to compete against.

At the age of 11, when skills and strength, and the other attributes required for sporting success are still developing, 10 months is an eternity. Think back to your own development, or better yet, to the development of your children, if you have them. Backyard games of catch or football or rugby or cricket change dramtically from one year to the next because a child at that age acquires skills and strength so quickly that they improve enormously from one month to the next.

This means that the 10-month advantage that Ross has, by virtue of being born in January or February, will manifest itself as a big performance advantage over Jonathan (obviously, I'm generalising here, you'll find exceptions. But the graphs above suggest that they happen infrequently).

Enter the coach, and the Matthew Effect is born

Now the coach enters the picture. He has a team of energetic, uncontrollable young 10-year olds to look after, and he picks his team, and allocates his time and attention to those who are deemed to have the most potential. However, he is unable to distinguish between capacity for performance and maturity. Maturity determines ability - Ross is older, and may possess more strength, speed, skill and therefore appear the star player in practice. Jonathan is yet to develop these attributes, but may have the talent.

However, the fateful decision made by the coach is to pick Ross ahead of Jonathan. What happens next determines the distribution you see in the graphs above. By virtue of having been picked based on his "superior" ability, Ross plays against higher quality competition, receives better coaching, more attention, and therefore improves MORE than Jonathan will. Their paths are determined from the outset, based on their selection, and the different journeys they will take are going to mean that one day, Ross is the better athlete or player, thanks to these advantages and opportunities he has received.

At the same time, Jonathan is far less likely to continue to play, because:

1. He is often smaller than the guys he competes against, and that's not likely to make his life much fun!
2. He becomes discouraged at the ever-growing gap between himself and the other guys who are being more heavily invested in

What we have then, is a self-fulfilling prophecy, where Ross is picked because he is better, and then ends up being better because he was picked, apparently vindicating the coach's early decision. The issue is this: Was his initial selection the result of his age, or was he genuinely the better sportsman? That is the challenge for talent identification. The figure below summarizes the process.

And why is it called the Matthew Effect, you may be asking? That name was coined by the sociologist Robert Merton, based on the bible verse from Matthew: "For everyone who has will be given more, and he will have an abundance. Whoever does not have, even what he has well be taken from him" (Matt 25:29).

That means that success comes to those that are successful, thanks to their advantage, in this case, from their relative age.

Confounders and debate

There are of course debates and issues around this. The presence of athletes who are born between September and December (albeit a low 13%) suggests that exceptions do exist. What might be very instructive is to examine how those "outliers" (apologies, Malcolm) reach the level they do - they may be early developers, they may have parents who start them out by playing games at a younger age developing their skills "ahead of the normal curve", or they may have older siblings who drag them to a higher level of performance despite their younger relative age. All these options are intriguing and instructive for talent ID purposes, and for understanding how sporting success is determined.

The three elements required: Selection, streaming and differentiated experience

To dig into the effect a little further, the afore-mentioned Roger Barnsley suggested that three things are required for this effect to occur:

1. Selection - someone (in this case, a coach) must be selecting players based on ability
2. Streaming - once selected, players are placed into streams. These can be competitions, teams, training squads etc.
3. Differentiated experiences - very importantly, once in the streams, players receive different levels of coaching, competition and opportunity. This is summarized in the figure above.

Now, all three of these characteristics are found in what is called a meritocracy - wherever performance is rewarded through selection based on ability, and the pyramid of progress exists, you will generate this skewed distribution. Where there is no meritocracy, the effect is diminished, because the emphasis on the selection of the best players (regardless of age) is not as prioritized.

The lack of a meritocracy can be seen in the slightly more balanced distribution of the SA schools rugby team, by the way. In that team, "only" 44% were born in the first three months of the year, much lower than for the other three teams. So what, you ask? Well, that's because the SA school team was selected out of a mix of "traditional" and "non-traditional" rugby schools. Without going into the history and politics of our SA rugby, we have some schools that are very heavily based around performance, and have multiple teams at each age group, right down to junior level. Others are not as focused on age-group performance, and because of the selection policies in SA sport, the team is a mixture of the two. This dilutes the Matthew Effect. You'll find that whenever merit is NOT the primary factor for selection, this will occur.

Implications for sports science

Well, I'm running out of time (and so are you, probably!). This post has run over-time, but there's much more discussion on this topic to come. One area that interests me in particular is that talent ID often does not note this potential confounder of age. It is for this reason that the identification of talent, especially if that result is going to be used to "stratify" young sportspeople into different paths, should occur as late as possible.

What this suggests is that a sports scientist who plans to do some talent ID on a group of unknown sports people might as well make his first selection on the basis of date of birth! OK, that's being too extreme, but the point is, by the time a sports scientist tests a squad of potential athletes aged say 19 years, it's often too late to undo the effect that he measures, and which may have been created 9 years earlier. More to the point, the sport in question might be missing out on some of its best potential talent, which was lost 9 years earlier thanks to a wrong selection! At the senior level, the coach who picks the team may believe he is picking from the whole pool, when in fact, 40% of it has been lost through premature selection, 9 years previously! That should be a significant "flag" for people involved in high performance sport...

More to come...

Mulling over the Matthew Effect

Yesterday, I did a post describing something that has been called The Matthew Effect, as applied to sports performance and talent Identification. Briefly, it refers to the phenomenon where a disproportionate number of elite level sports people are born during the first few months of the year. This comes about as a result of a confusion between ability and maturity, and the selection of those children into regional or school teams based on their ability at a young age. Unfortunately, at this age, 10 months is a significant difference, and those who happen to be younger are soon left behind.

In response to the post, we received some really interesting and thoughtful comments, which you can read at the original article. I tried to respond to those comments, but it's worthwhile mulling over a few of the points raised. In particular, the big question is: Given this effect, and the fact that children born in later months are seemingly disadvantaged by their younger relative age, what should be done to ensure maximum talent "delivery" at the senior level?

The key point is that a scientist who is faced with the seemingly daunting task of selecting a squad of sportsmen from a completely random sample would be reasonably accurate simply by asking everyone born in January, February and March to stay behind, and sending home those born in the latter part of the year! This in turn means that all those who are born later in the year are placed at a disadvantage, and you create a vast pool of "unrealised potential".

It's important to note that by the time the national coach, or the head coach of a profession team makes his/her picks, it's already too late. The damage has been done, many years earlier. Similarly, a sports scientist who is doing talent ID assessments for a high performance programme at say Olympic level cannot be concerned with month of birth - their job is simply to pull out those people who display talent or ability to perform better than others. The problem is, most of those they pull out will have been born early in the year, thanks to a decision made many years earlier. The question is what one should do to ensure that up to half the population remain eligible for success for as long as possible?

A split in age-groupings?

And that is a question I was thinking about a great deal today, and must confess that no easy solution presents itself. In his book, Gladwell suggests the creation of a "split" in the age groupings, so that children born in the first half of the year compete in a separate league structure to those born in the second half. I'm not sure about ice-hockey (which is the example he uses), but this idea would be very difficult to implement for most other sports. For one thing, it would require twice the coaching time and expertise and would more than double some of the resources required for the participation of athletes in sport.

That is, the entire basis for the Matthew effect is that younger players who happen to be older by virtue of their earlier birth month are given superior coaching, competition and opportunity. The creation of a second, separate league for those born later in the year would not solve this problem, unless the quality of coaching provided to that second group of children was at least comparable. In SA, there are barely enough decent coaches for one team, let alone two, so I'm not convinced this would work. And that's apart from the administrative problem raised by Gladwell in his book.

Weight categories?

The second possibility is the creation of weight categories for young children. The rationale here would be that in sports where size, weight and strength (these are often, but not always, associated) are key determinants of performance, the early developers and the relatively older children enjoy a large advantage which manifests itself as improved ability, and which is the basis for the fateful selection of January births rather than December births.

The creation of weight limits would ensure that children only compete against those who are in the same weight bracket as they are, regardless of age. It has some advantage, but there are also a couple of problems. The first is the incentive it creates for children to make weight. That's not to suggest that the use of anabolic agents or diuretics (depending on which direction they wish to go) would be the obvious result, but it is a possibility.

More than this, it creates something of a perverse criteria against which children are measured, one which I'm not sure is healthy. It also starts to mix children of very different ages together, and there is an emotional and intellectual difference that is not controlled for. Suddenly a 10-year old is playing sport against 15 year olds, in a league that is, by nature, likely to be much more competitive than should be the case for a 10-year old.

Secondly, many sports actually require a separation in weight before specific skills can be acquired. Being South African, the sport that comes to mind is rugby, but for those in the USA, the obvious one might be American Football. In both sports, positions are very heavily influenced by body shape, size, strength and physical stature. As a result, so too are the skills required from the players in those positions. If players compete in age categories, then one would be delaying the acquisition of these skills.

Is that a bad thing? Not necessarily. Skills develop according to the situation presented to the player, and so a more all-round skill set would be the result if age-categories were adhered to, since no size advantage would exist for one player to easily dominate another physically. Late maturers would be rewarded, because they would develop skills that one would usually be found in the smaller, more nible players, and when they eventually "fill out" and bulk up, they'd carry through those skills, and remain skillful as "big men".

On the other hand, do you really want a team with all-round skills and reduced specialist ability? Once you reach the professional level, the very specific demands of playing each position would quickly expose weaknesses that have developed as a result of the lack of necessity to develop those skills earlier. It's a debatable one, for sure. It has been tried, that much I know - I believe that they tried weight groupings (mixed with age groupings) in Australia. I haven't had the chance to investigate that more thoroughly, so I'm open to input on that one.

Changing the focus of performance

Perhaps the best approach I can think of borrows from this principle that you want to discourage a form of play where size, strength and speed are the crucial factors that determine success. In this regard, many sports systems around the world are already making the effort, since they emphasize that younger children do not play contact forms of sport, play rather for fun and enjoyment and do not prioritze winning. The notion of play to play, rather than play to win, is the focus.

Recall that the Matthew effect develops when coaches select players and then begin to provide a superior coaching and competition environment. If those coaches are able to make their selection in such a way that ability is not confused with maturity, then younger players would remain in the system for longer, perhaps long enough that they could themselves develop and catch up to the older players.

The incentive (and the wisdom) of the coach is therefore the first element - they should not be driven to win. Unfortunately, in many sports, this is an unrealistic goal, and one can understand how coaches pick better players at such young ages - they are under pressure to perform. So the collective mindset of the team, the parents, the school and the club often must change before this happens.

Once it does, then the priority of the coach can become skill development, enjoyment and development of attributes where size, strength and speed are not solely responsible for performance. This will never completely remove the effect, but education, a change in mindset and a different set of priorities might go a long way to "rescuing" those younger players who are so quickly lost from the system. Ironically enough, this focus on play rather than performance at a young age is likely to help performance at an older age, through the creation of a larger body of "eligible" players.

An impossible puzzle to solve, I suspect. As I said, I'm actually going to be suggesting to a few sports federations here in SA that they look long and hard at this very phenomenon, and try to understand how young talent moves through the system. Part of this will be discovering where these "early developers" go, what happens to late developers, and what strategies might be effective in maximizing the available player pool. So we have the possiblity of a real-life "case-study" or two, and I hope to be able to report on that soon!