This article is shared by Player Development Project
Steve Lawrence is a consultant to Cruyff Football and Ajax in the Netherlands. With a Masters in International Sports Management, Steve founded the Football Analytics Lab and is at the forefront of research into the topic of Relative Age Effects.
In this article, Steve explains how the Relative Age Effect works and discusses the impact of this phenomenon.
Relative Age Effects have become a well known, if slightly puzzling, phenomenon in youth football where players with birthdates at the beginning of the season have a huge advantage over their teammates born later in the season.
Relative age effects are not confined to football but can be seen across sport, education and childhood development generally. Hundreds of academic papers have been published on relative age, birthdate effects and school entry age effects since the 1930s, but the paper which pinpointed the issue in respect of competitive youth sport was written in 1985 by Barnsley, Thompson & Barnsley and was titled ‘Hockey success and birth-date: The relative age effect.’
A slightly later paper by Barnsley & Thompson in 1988, also about ice-hockey, went on to identify a ‘relative age advantage’ enjoyed by those with early birthdates, a ‘relative age disadvantage’ experienced by those with late birthdates and a ‘relative age difference’ between the two.
It is the ‘relative age difference’ which is the first thing to be clear about. It’s not the absolute birth month or time of year which is important but the proximity of a birthdate to any relevant cut-off date used for grouping children together:
- If a child’s birthday falls before a cut-off date they are excluded from a group.
- If their birthdate falls after the cut-off date they are included in the group.
This banding tends to be on an annual basis, and for academic purposes the cut-off date has historically been set in the autumn – which in the northern hemisphere has meant August or September.
With increasing international harmonisation, 1 September has become an almost universal cut-off start date for banding children into academic year groups – and so it was in football until 1997.
Between the 1995 and 1997 youth World Cups, FIFA decided to change the cut-off eligibility date for the U17 and U20 World Cup competitions. The cut-off date was moved from 1 August to 1 January and circulars were sent out to member Confederations and Associations. Most of world football subsequently adopted 1 January as the administrative cut-off date, the FA in England being a notable exception where a cut-off date of 1 September still applies.
“In this system the ‘relative age difference’ between a child born at 23.59 hrs on 31 December and a child born on at 00.00 hrs on 1 January is not one minute, it is one year – and that’s where the problems begin.”
To all intents and purposes, 1 January is now the globally accepted eligibility cut-off date for children’s football and for sport generally. This is not true for academic education, however, where 1 September remains the global standard.
So the ‘relative age difference’ in sport must nearly always be thought about in relation to the single calendar date of 1 January. Youth players born on or after 1 January enjoy ‘relative age advantage’ and those born in December experience ‘relative age disadvantage’. In this system the ‘relative age difference’ between a child born at 23.59 hrs on 31 December and a child born on at 00.00 hrs on 1 January is not one minute, it is one year – and that’s where the problems begin.
Whilst the setting of a single unique cut-off date for eligibility might be the cause and ‘relative age effects’ might be the consequence, the causal links between the two are both complex and contested.
One thing we do know is that relative age effects start to be seen at a very early age and they continue into adulthood.
The most understandable ‘relative age effect’ and the one which is most familiar is the very noticeable bias in participation and selection, particularly at higher achievement levels.
It’s obvious that sporting talent is not confined just to those born in a particular month and we ought to expect to see a natural distribution of talented players across the whole year, reflecting the normal birth distribution which looks like this:
However, if we look at a dataset of players who see themselves as talented, we see this:
The graph shows a peer-selected sample of 200,672 players. Their peers, parents and trainers post their details on the soccertalents.com website because they see them as ‘talented’.
By age 7 the twin peaks have disappeared and January remains as the single peak, with 14% of ‘talents’ having January birthdays and 3% of ‘talents’ having December birthdays.
At age 5, which is the age at which talents start to have details posted, we see two peaks, one in September coinciding with starting at school and one in January coinciding with joining organised teams. By age 7 the twin peaks have disappeared and January remains as the single peak, with 14% of ‘talents’ having January birthdays and 3% of ‘talents’ having December birthdays. Many soccer schools use the 7th birthday as the eligibility date for entry with insurance policies and so on drafted on that basis.
The eagle eyed will notice a September blip lasting until around age 13 and another at age 25. The later blip is to do with the FIFA rule change in 1997 and the early blip is probably the continuing influence of school entry and/or English players.
So we can see that the early influence of cut-off eligibility rules is to set up a bias between those who see themselves as proficient and those who don’t based on their relative age rather than their actual relative proficiency. That self-image is reinforced by competitive results because relatively older players are more successful than relatively younger players. The relatively older players gravitate to the successful teams, which in turn are increasingly successful because they are populated by relatively older players. The relatively younger players are excluded from the more successful teams and see themselves as less successful and therefore less talented.
“Unsurprisingly they recruit relatively older players as candidates for specialised training and inclusion in the more elite levels of youth football.”
The next step in reinforcing this ‘relative age effect’ is promoted by scouts and coaches who are on the lookout for successful players in successful teams, and unsurprisingly they recruit relatively older players as candidates for specialised training and inclusion in the more elite levels of youth football.
In an experiment amongst amateur clubs around Amsterdam in 2013–14, we asked the trainers at 15 clubs to select their best squads from a total pool of 3,120 players at the ages of U9, U10, U11, U12, U13 and U14. We explained the relative age effect in advance to the trainers so they would have it in mind with the suggestion that they should select on the basis of talent and not age.
Their selections are shown on the graph with the percentage of early born players (born January-June) expressed as an integer between 0-1. I refer to this as the relative age effect index or RAEi – it just gives a rough guide to the amount of bias in any group.
The blue data points are the selected teams for higher level competition (1,036 players) and the greens are the remainder chosen to play at lower levels (2,084 players) and the selection process is clearly biased. It’s worth noting that the data points have a degree of reflective symmetry because, of course, this is a zero sum exercise – an early-born player chosen for the selection squad means a late-born player has to go into the non-selection squad.
Also plotted on the graph are the Ajax youth teams at U13, U14, U15, U16 and U17, along with the data points from the soccertalents database above as a reference.
So we see more early-born players being seen as ‘more talented’ by their peers and importantly by coaches and scouts. They are therefore selected for the squads which receive more attention, training, equipment (the selected players get new kit whilst the remainder make do with last season’s) and the Academy club scouts then come and watch the selection team matches (why would they bother with the others?) and lo and behold an even more biased group ends up in the Academy teams, where yet more advantages accrue.
It’s an upward spiral of advantage for the early-born player who is recognised as talented, while it’s a fight for survival for the late-born player.
The slow but sure decline in relative age bias evident from the soccertalents dataset gives a clue to eventual stability, with no bias at around the age of 28.
If we look at some data from the Premier League in 2013–14 this is confirmed and it allows us to set up a straight-line model which is a pretty good rule of thumb for calculating the relative age bias in high performing teams from about 90/10 at age 6 to about 50/50 at the age of 28.
This age of parity is very interesting and I explored it in a paper called ‘The Age Advantage in Association Football’. The premise being that there is a tipping point such that a team, with an average age close to that tipping point age, will tend to beat another team with an average age further away from the tipping point age. I refer to this as the ‘optimum’ average team age or ‘optimum ATA’.
This optimum average team age is around 27.3 – which coincides with the age at which we see relative age bias in high performing teams diminish to around 50/50. This, of course, makes sense because in any ruthlessly efficient system waste is abhorred, and in our case attrition of early-born but less able players takes place, reducing their numbers to match the late-born ‘whispering talents’ who have remained but whose prowess goes largely unrecognised until they begin to approach the optimum age.
“How is it possible to recognise late-born ‘whispering talents’ and, even more difficult to answer, how can they be sustained until they reach 23 or 24?”
All of this gives rise to a very perplexing situation for trainers and coaches: how is it possible to recognise late-born ‘whispering talents’ and, even more difficult to answer, how can they be sustained until they reach 23 or 24? At the same time, how is it possible to recognise the early-born likely underachievers on whom development resources mustn’t be wasted?
The common suggestions include coach education, quotas, parallel competitions, multi-age grouping and so on – but my view is that these actions can only ever scratch away at the problem, they can never solve it.
The competitive nature of football is such that coaches and trainers will tend to field teams that are likely to win matches. At adult ages, this means that teams will tend towards the optimum, and where cut-off eligibility dates are in force that means fielding teams where the average team age is as close as possible to that cut-off date. And that means fielding teams with relatively older players, which results in a high relative age bias.
So if a cause has undesirable and unintended effects then why not remove it? It may be a challenge if the system has become universal, but that doesn’t mean it’s not soluble. I would even go as far as to say that those who look for and adopt a solution to relative age bias will benefit by being first to clear a wasteful use of resources from their recruitment and development system.
My feeling is that the solution to the problem is crystal clear and elegant: to simply remove the opportunity for one team to be significantly older than another and an answer lies in the optimum average team age analysis.
So what does an optimum profile look like? I’ve been developing ideas for a hypothetical normal profile with Simon Gleave, Head of Analytics at Infostrada Sports, and whilst this is work in progress an optimum curve might look something like this:
The age curve for the starting line-up for Germany in the 2014 FIFA World Cup Final makes an interesting comparison with the hypothetical optimum, with the late bump being Miroslav Klose who wasn’t on the field at the end of the match having been substituted for Mario Gotze, the goal scorer.
There was also no relative age bias in the German team – the ratio was 45/55, so as close as possible to 50/50 with 11 players.
For me this raises the question that if we don’t see any bias at the peak performance ages surely we should look to establish a system of recruitment and player development which has no bias at any age?
We would then know that the system was at its most efficient with the most talented individuals involved at any given time and with resources used most efficiently.
“The answer is quite simply to change the eligibility rule”
So how do we remove the opportunity (in youth football) for older teams to line up against younger teams? The answer is quite simply to change the eligibility rule from a ‘cut-off date’ for individuals to a rule which requires both teams to meet an ‘average team age’ eligibility rule so that both teams are broadly the same age.
By introducing a banding criteria which says that the oldest player in a team or squad should be no more than say 2 years older than the youngest player in the squad, a kind of bio-banding solution is also achieved which allows coaches great flexibility in devising squads to allow individuals development opportunities within squads.
The consequence of such a rule would be to create an advantage for talented younger players to play in slightly older age groups because they would bring the average age down to meet the criteria, creating a system in which the cream would tend to float to the top. This would in turn change the emphasis in youth football from ‘talent identification’ to ‘talent emergence’ and I would argue that that would be a very good thing for the game.
Image Credit: Photo Sport NZ