How Fast Do PSA Populations Actually Grow?

A population figure is a stock. What matters to anybody holding a slab is the flow, and almost nobody publishes it, because you need the same registry read on two different days. We have been recording PSA registry snapshots daily, so we can answer it for a whole population rather than a handful of famous cards. Across the 2,916 registries we watched for at least 28 days to September 22, 2026, a median 39 days each, 24.93% added no slabs at all and the busiest 1% absorbed 29.87% of the 376,981 slabs that were added. Grading is not a tide. It is a spotlight.

This is the part most growth claims skip. Our registry tape is 51 days deep at its deepest, and cards enter it at different times, so the set of registries with a 14-day window is a much larger and much shallower population than the set with a 42-day one. The share that appears frozen falls from 65.86% at a seven-day window to 5.86% at a forty-two-day one, not because grading sped up but because a registry adding a slab every few weeks looks dead over seven days and alive over six. We publish the whole ladder and take the 28-day cohort as the headline, because it is the shortest window at which the median observed span is over a month.

Two things survive the whole ladder. The concentration never drops below a sixth of all new slabs landing on 1% of registries, and at the shorter windows it passes half. And a large minority of registries are simply not moving at any window you pick.

In the headline cohort the median registry added 13 slabs over its window, which is 0.26% a week. The 75th percentile is 0.50% a week and the 90th is 0.86%. Only the top 1% run faster than 6.31% a week. Set against that, the premium decay people worry about is slow for almost every card and violent for a few dozen.

One warning about that last row before anybody quotes it. The fastest-growing registry in the cohort by percentage is Morpeko ex from M5: Abyss Eye at 63.87% a week, and it earned that by going from 71 slabs to 369 over 46 days. A percentage on a base of 71 is a statement about the base. Restrict to registries that opened above 1,000 slabs and the fastest is Chimchar from MEP Black Star Promos at 18.78% a week, 4,686 to 8,708 in 32 days, which is a genuinely fast registry rather than an arithmetic artefact.

Size decides whether a registry moves at all. Half of the registries under 500 slabs added nothing across their window, against 0.79% of those between 10,000 and 50,000. But the growth RATE barely varies with size once a registry is moving, sitting between 0.28% and 0.41% a week across every band above 500. Big registries add more slabs because they are big, not because they are growing faster.

The Mega era is the fastest-growing block on our tape by a distance, at a median 1.28% a week against 0.16% for Sword & Shield era cards, and it did that from only 70 registries. Japanese registries in the cohort are both more likely to be moving, 14.64% frozen against 27.38% for English, and much busier per registry, a median 45 slabs added against 9. That matches what japanese slabs found about registry depth and what where the new slabs go found about submission mix.

Eight of those nine cards come from sets released in the last eighteen months. The one exception, the M-P promo Pikachu, is the deepest registry in our data and it grew at 0.53% a week. New product is what the grading rooms are full of, which is the supply side of the mechanism the grading throughput index measured from the aggregate direction.

Intersect this cohort with the cards that also carry a raw ask, a PSA 10 sold median on ten or more comps and an opening registry of 500 slabs or more, and 1,228 remain. That last gate matters and is not free: it drops the small registries whose premiums are widest, and without it 1,413 cards qualify and the three medians run 11.71x, 9.08x and 8.12x instead. Sorting them by weekly registry growth, the slowest third carries a median 9.57x premium and the fastest third 7.05x. The rank correlation is -0.16, which is real and small. Over 39 days the flow is a light drag on the premium, not the dominant force; the level of the registry, which does a big PSA population lower the price measures at -0.46, matters roughly three times more.

The deepest window here is 51 days and the median is 39, so this is one autumn and not a law. A set that launched in August is over-represented in the fast tail and a set launching in November is absent from it entirely. The cohort is also self-selected by our own tracking: a registry enters when we start reading it, and we read the cards people look at, so the sleepiest corners of the catalog are under-counted here rather than proven still.

A registry that added nothing is not proof that nobody submitted. PSA publishes a snapshot and the provider serves it to us; a registry can sit unchanged because no slabs were minted, or because the line was not refreshed in the window we watched. We cannot separate those two, and the honest reading of 24.93% is an upper bound on how much of the catalog has genuinely stopped. Registry totals are lifetime cumulative counts that include resubmissions and crossovers, so a slab added is not necessarily a card newly graded. And the continuity rule cuts a series wherever a reading contradicts population physics, so a card whose provider line changed identity mid-window contributes only its current line.

Re-measure the same 2,916 registries over the next 28 days. If a clear majority of them grow, or if the busiest 1% absorb under 10% of the new slabs, the concentration finding is wrong. A weaker test is available immediately: the growth rate should stay roughly flat across registry size bands above 500 slabs while the frozen share collapses with size. A band above 500 slabs whose median weekly growth is double another's would break the description. Related: the population growth tax, which ran this question on 13 cards over 17 days, and the registry is closed.

Source: `population_snapshots`, our own daily recording of the PSA registry line served for each tracked product, read through September 22, 2026. Each registry's series is passed through the house continuity rule, which cuts the series at any reading that contradicts population physics, a total that fell or a step that multiplied it threefold, because those are provider remaps onto a different registry line and not a week of grading; only the segment since the last cut is measured, and a line whose identity just changed with a single reading behind it is withheld entirely. Keys flagged by the shared-line rule, where two products return the same registry line, are dropped. Rows are collapsed to one per registry key. The headline cohort is every surviving registry whose publishable segment spans at least 28 days with at least two readings and a positive opening total, which is 2,916 registries with a median observed span of 39 days and a maximum of 51. Growth is closing total minus opening total, expressed per week as (added / days) x 7 and as a percentage of the opening total. The article publishes the same measurements at five other minimum windows so the reader can see how much the window choice is doing. The price join uses the latest `price_history` ask and the latest PSA 10 `graded_snapshots` sold median on ten or more verified comps, collapsed to one row per product id, and additionally requires an opening registry of 500 slabs or more, because a weekly growth rate computed on an opening total of forty is a statement about the forty; the article prints the result without that floor as well. Raw figures are asks, graded figures are sold medians. Not financial advice.