Does a block of altitude training precede faster 100 m sprint times?

The question

Altitude blocks are among the few training decisions in the sport that are expensive, scheduled a year ahead, and taken on a coach’s judgement rather than a published estimate. That combination is what makes the question worth putting to the record: a programme committing three weeks and a travel budget to a camp has a reasonable claim on knowing what the record of previous camps shows.

The question is narrow on purpose. Not whether altitude training works, which is a physiological question this record cannot address. Not whether a particular camp helped a particular swimmer, which no cohort comparison can support. Only this: in the record Swimetrica holds, did swimmer-seasons containing an altitude block of 14 days or longer show a larger improvement in adjusted 100 m time than comparable seasons without one?

What was compared

Every race in the window is placed on one scale by the age–sex–pool adjustment v3.0. A swimmer-season’s improvement is the change in adjusted time between its first and last quarter. An altitude block is a training camp of 14 days or longer at 1 800 m or higher, as recorded by the supplying federation’s camp register.

The raw comparison is between 641 swimmer-seasons that contain such a block and 2 477 that do not — and it is not the article’s estimate. Swimmers who attend altitude camps are selected into them: they tend to be older, in funded programmes, and improving faster before the camp than the swimmers they are being compared with. The matched comparison pairs each altitude season with control seasons on prior improvement rate, age at the start of the season, event, sex category as recorded, and programme. 1 839 of 3 118 eligible seasons matched inside the calipers; the rest are reported as not matched rather than dropped.

Result

The raw difference, before matching, is −0.21 s (95% interval −0.28 to −0.14), which excludes zero. After matching, the estimated difference is −0.03 s (95% interval −0.11 to +0.05), which includes zero. The gap between those two figures is the size of the selection into altitude camps in this record. The takeaway: the raw comparison — the one most easily made from a public results database — mostly reflects who goes to altitude, not what altitude does for them.

Six event-and-sex subgroups were estimated separately and every one of the six intervals includes zero:

The narrowest interval is men’s 100 m freestyle at 0.28 s wide, on 508 seasons; the widest is women’s 100 m butterfly at 0.42 s wide, on 188. An interval that wide is compatible with a difference of a tenth of a second in either direction — a margin that decides races. These estimates do not distinguish an altitude effect from no effect, and they do not distinguish no effect from a large one.

What this does not support

• That altitude training has no effect. An interval that includes zero includes a great deal besides zero. The matched estimate is compatible with the altitude group improving 0.11 s more and with it improving 0.05 s less.
• That altitude training causes the raw gap. Matching removes the imbalance on the five variables it was given. It says nothing about an imbalance on a variable nobody recorded — funding, physiotherapy access, hours in the water, or a coach’s judgement about who is well enough to travel.
• That this result transfers to one swimmer. A cohort comparison is a statement about a distribution of seasons and carries no estimate for any individual.
• That this result transfers to another event, distance or age band. The comparison is three 100 m events, long course, in swimmers 18 and over.

Method and limitations

Six limitations bear directly on how wide these intervals should be read. The camp register is a record, not a measurement: a camp that happened and was not recorded enters this analysis as a control season, which biases the estimate toward zero by an amount this article cannot quantify. Coverage of the register is uneven — nine of the 19 federations record both dates and altitude, six record dates only and enter the comparison on the 14-day threshold with altitude assumed from the venue, and four record neither and are excluded entirely, removing 431 seasons from the eligible population. Matching is not randomisation: the calipers were 0.15 s per season on prior improvement rate and two years on age, and the 1 279 seasons outside them were not matched. A sensitivity check at wider calipers moves the pooled estimate to −0.05 s (95% interval −0.12 to +0.02). One season is one observation and swimmers appear more than once in the window; the interval accounts for that within-swimmer correlation, and a specification that did not would report an interval roughly a third narrower. The improvement measure inherits the age–sex–pool adjustment’s own calibration, which is 94.2% against a nominal 95% (method: how the intervals are calibrated). Swimmers under 18 are outside this population by policy, not by data availability.

What would narrow the interval

Three things, in order of how cheaply they could be had rather than how much they would help. Coverage: a register that recorded venue altitude directly would move roughly 400 seasons from an assumed exposure to a recorded one. Duration: the 14-day threshold is a convention borrowed from the physiological literature, and this record contains blocks from 8 to 34 days; estimating against duration rather than a threshold uses information the threshold discards, at the cost of a stronger functional assumption. Time: the window contains seven seasons; at the current rate of accrual the pooled interval reaches roughly 0.11 s wide by the 2030 season — still wide enough to contain the margin most readers care about.