Aalen–Johansen · three states · monthly
Getting out of a division
Every dancer who reaches a WSDC division either climbs out of it or stops competing. This is how that race runs, one rung at a time — measured from the month the ceiling rose into the division, and followed for ten years.
Retirement is a competing event here, not censoring, so advanced, stopped competing and still climbing sum to 1 at every month. Censoring a dancer who quit would assume they were going to advance at the same rate as one who stayed.
The rug shares the axis above it. Each mark is a month in which dancers left observation still un-advanced and still competing — censored, not retired. Censoring happens in nearly every month here, so the usual tick-on-the-line convention would draw a second solid line over the first; the rug keeps the shape instead, and the exact counts are in the table below.
Where everyone ends up
The same cohorts, decomposed. The three bands account for the whole population of the rung at every month, so what each panel answers is not "how fast do people advance" but "of everyone who was here, where are they now".
How much of this is the cutoff?
"Stopped competing" is a heuristic: a dancer with no scoring placement for 25 months reads as retired. Re-running the estimate at 12 and 36 months shows how much of the headline belongs to the rule rather than to the dancers.
Read this as the width of the answer. The 25-month cutoff is the one every other number on this page uses. The spread between the 12- and 36-month runs is neither error nor noise — it is the part of "how many advance" that the cutoff decides.
What these numbers are not
Table view — every plotted month, every division
Rows step through the plotted range, 0–120 months. The terminal row carries the end of the full curve, which runs further than the chart does — out to month , on a risk set of one.