Junior champions

Why do the children who win national age-group titles so rarely become the adults who win senior ones?

A national under-15 title looks like a measurement. Someone ran the race, the clock was honest, the child on the top step really was the fastest in the country that day. And yet follow those top steps for a decade and the conversion rate into senior champions is disappointing enough that federations keep commissioning studies about it.

So the awkward question is not "why do juniors fail?" — most of them do not fail at anything. It is: what quantity was the junior podium actually measuring, and is that quantity the same one the senior podium measures?

Begin with what a race at fourteen can possibly detect. It records today's performance, which is some combination of three things: whatever trainable capacity the child has, how much training they have already done, and how far through puberty their body currently is. The clock cannot separate those. It returns one number.

Now ask how much the third term varies inside a single birth-year cohort. This is the part that surprises people. Among children of the same chronological age around thirteen or fourteen, biological maturity spans nearly the whole range there is — some are still entirely pre-pubertal, others essentially finished. Puberty in boys adds stature, lean mass, testosterone-driven muscle and a rise in haemoglobin; the effect on any event that rewards power, strength or oxygen transport is not a small edge but a category difference. So in a maturity-sensitive event, ranking a cohort by performance ranks it largely by maturity, with talent riding on top as a smaller term.

Here is the move that makes the whole puzzle dissolve. That advantage has a date on it. Everyone matures eventually. The variable the selection leaned on is one that expires — by twenty, the early maturer holds none of it, because the late maturer has arrived at the same place. Whatever ordering the junior podium produced, the ground under it is being removed year by year.

Does anything else contribute? Two things, and they pull in opposite directions. Some of any single result is noise, so an exceptional performance tends to be followed by a more ordinary one for purely statistical reasons — but notice that this is a different mechanism from the one above, and worth keeping separate. Statistical regression is symmetric and undated: it applies to a fluke result at any age. Maturation is systematic and expiring: it predictably favours one identifiable group and then predictably stops. In the other direction, being identified early brings coaching, squad places and competition, which is a real advantage that partly makes the selection true after the fact.

And the child on the other side of that filter? A late maturer competing against half-grown adults loses, gets less coaching, is told less often that they are promising, and is disproportionately likely to leave the sport before their own body arrives. That is not a story about the winner regressing. It is a story about the sport not having measured a large part of its own cohort at all.

Three refinements, because the claim is easy to overstate.

First, a junior title is not evidence against an athlete. The population claim is that conversion is much lower than the podium's apparent authority suggests — not that early winners are handicapped. Any individual junior champion may be genuinely exceptional and go on to win everything; the statement is about a rate across a cohort, and it says very little about the child in front of you.

Second, the size of the effect should depend on how maturity-sensitive the event is, and it does appear to. This is what makes the account testable rather than just plausible.

Third, honestly: separating the three contributions — expiring maturity advantage, statistical regression, and differential dropout — is genuinely hard, and researchers argue about the weighting. The reason is structural. The children who were deselected are largely unobserved thereafter, so the very data that would settle the split is the data the selection destroyed.

What would falsify the maturation story? Measure biological maturity directly in age-group finalists — skeletal age, or years from peak height velocity — rather than inferring it. The account predicts finalists in power and endurance events are systematically advanced relative to their cohort. If their maturity distribution simply matched the cohort's, the mechanism would be dead and regression plus dropout would have to carry the whole load. A sharper test comes from events where late maturation helps: in disciplines that reward a small, light, still-childlike body, the filter should select the opposite maturity type, and so should behave differently — if maturity mattered, its direction should flip with the demand.

• How this differs from the relative age effect, which turns on birth date within the band rather than on biological maturity — and why bio-banding addresses one and not fully the other.
• Whether the athletes lost to early deselection can be recovered, or whether the missed training years are themselves decisive.

Every term the collection defines is gathered in the glossary.