One card, one cycle, one catastrophe. The same demand shock produces opposite outcomes depending on a single variable, and each case ends with the thing a model would not have known in advance.
Thesis
Whether a price move persists depends less on how strong the demand was than on whether supply could respond to it. Read the three episodes in that order and the outcomes stop looking like sentiment and start looking like elasticity.
Abstract
Three cases, chosen because each isolates a different mechanism. Base Set Charizard shows demand and grade convexity: roughly 4,993 copies certified, about 124 of them PSA 10[1], and a public record that moved from $550,000 to $1.7m in fifteen months[2]. The 2020–22 cycle shows a demand shock meeting a market where only half of supply could respond: modern fell 30–50% from peak while vintage quality material held[3]. The junk wax era shows what happens when supply answers without limit. Output roughly tripled, and Topps alone printed an estimated billion cards in 1986, and most of that era is now worth cents[4]. Each section closes with what a model would have needed to anticipate the outcome, and what it could not have known.
How to read these
A case study that only explains what happened is a story, and stories about markets are cheap. Every price path has a narrative available after the fact. Each of these therefore ends with two harder questions: what a model would have needed to observe to anticipate the outcome, and what it could not have known at the time. The second question is the useful one.
Case A: Base Set Charizard
The 1999 Base Set Charizard is the most recognisable trading card of the modern era and a clean demonstration that rarity and scarcity are not the same thing.
Roughly 4,993 copies of the 1st edition have been certified. About 124 of them, under three per cent, carry a PSA 10[1]. The card is not rare in any ordinary sense: five thousand of anything is a lot. What is scarce is the top of the grade ladder, and the ladder is where almost all the value sits.
Where the scarcity actually is
Figure 1
- 2% PSA 10 · 124
- 98% PSA 9 and below · 4,869
Roughly 4,993 copies certified, of which about 124 are PSA 10. The card is not rare. The grade is.
Source: [1]. Population as reported. A census only grows, so this is a floor, not a fixed number.
This is grade convexity, and it is the single most important structural feature of graded collectibles. The gap between a 9 and a 10 is not one step of condition; it is the difference between an asset with five thousand substitutes and one with a hundred. Any valuation model that treats grade as a linear input will be wrong about exactly the cards that matter.
The price path over the last fifteen months is correspondingly steep. A PSA 10 1st edition set a public auction record at $550,000 in December 2025; a reported sale reached $954,800 in February 2026; and in March 2026 a PSA 10 of the Japanese Base Set printing set an all-time record for the character at $1.7m[2].
Public records, fifteen months apart
Figure 2
Source: [2].
Which clock was running
Demand, almost entirely. The people who opened Base Set packs in 1999 were children; they are now in their thirties and forties. Nothing about the card changed. The supply of PSA 10s grew only as fast as anyone found another gradeable copy, which is very slowly.
What a model would have needed
- —Population at grade and above it, tracked over time rather than as a static number, to establish that top-grade supply was effectively fixed.
- —A convex grade term, because a linear one would have priced the 10 as a slightly better 9 and missed the entire effect.
- —A wide interval. A card trading a few times a year cannot be valued precisely, and a model claiming otherwise is the dangerous kind.
What it could not have known
That a Japanese printing would set the character's record. Cross-printing substitution, collectors moving between the English and Japanese versions of the same card, is a demand shift with no observable leading indicator in any population or price series. A model would have been holding the English card and watching a different one reprice.
Case B: the 2020–22 cycle
The boom is well documented: card sales on eBay rose 142% in the boom year, basketball 373%, soccer 1,586%[3]. The correction that followed is the more instructive half, because it did not fall evenly.
Modern and ultra-modern cards gave back somewhere between 30% and 50% from their peaks, with tracked modern down over 30% in 2023 alone. Vintage material in quality grades held, with reports of no meaningful pullback in the Ruths, Mantles and Mayses[3].
The correction, by whether supply could answer
Figure 3
Source: [3]. Reported peak-to-trough ranges.
Which clock was running
Demand on the way up, supply on the way down. The same shock hit both halves of the market; only one half could answer it. A 2021 rookie card can be printed again and more copies can be pulled and graded. A 1952 Mantle in a high grade cannot be manufactured, and the pool of raw copies still capable of grading well is small and shrinking.
The usual explanation, that speculators left modern and collectors stayed in vintage, describes the same facts in terms of intent rather than mechanism. Supply elasticity explains it without needing to know what anyone was thinking.
What a model would have needed
- —A supply-elasticity term, distinguishing assets whose certified population can expand from those whose cannot. Population size alone is not enough: what matters is how much ungraded supply still exists behind the certified count.
- —Segment-specific parameters. A single model fitted across both halves would have averaged a boom and a contraction and reported something true of neither.
What it could not have known
The timing, and the size of the demand shock itself. Nothing in any price or population series in 2019 implied what 2020 would do to discretionary time and income. A model with a perfect supply term would still have been surprised by the arrival. It would only have been better prepared for the departure.
Case C: the junk wax era
Between roughly 1987 and 1994, manufacturers met booming demand by tripling output. Topps alone is estimated to have printed over a billion cards in 1986[4].
What tripling output looks like
Figure 4
Source: [4]. Only the 1986 Topps figure is a reported estimate. Manufacturers never published annual production, so the industry figures are ranges describing the shape of the expansion.
The outcome is the category's only genuine catastrophe. The vast majority of cards printed in that window are now worth cents, and a generation of collectors who had bought on the reasonable assumption that popularity implied value were left with boxes of them.
The detail that matters: demand did not fall. Interest in the hobby kept rising through most of the period. The collapse was entirely on the supply side, and it happened while every visible indicator of demand looked healthy.
Which clock was running
Supply, without a limiter. The feedback loop that normally restrains this, where higher prices provoke more supply, which lowers prices, which cuts supply, failed because manufacturers were responding to demand rather than to price, and could expand output faster than the market could absorb it.
What a model would have needed
Print-run data, which was never published. This is worth sitting with: the single most consequential variable in the category's history was one nobody outside the manufacturers could observe. A model built on prices and population reports alone would have seen a healthy market with rising participation right up until it was not.
What it could not have known, and what this implies now
The modern analogue does not require anyone to print anything. Grading converts an ungraded card into a newly tradeable unit, so certified supply can expand sharply with no new manufacturing at all, and unlike print runs this expansion is observable. 26.8 million cards were certified in 2025, a 32% rise[5].
That is the one genuinely actionable conclusion across these three cases. The variable that destroyed the category in the 1990s was invisible. Its modern equivalent is published annually, and is mostly being read as a sign of a healthy market rather than as a supply figure. We argue the case for treating it as the latter in The clock nobody watches.
What the three have in common
| Case | Dominant clock | Observable in advance? | What was missed |
|---|---|---|---|
| Base Set Charizard | Demand | Partly. The grade distribution was public | Substitution into a different printing |
| 2020–22 cycle | Demand up, supply down | The asymmetry, yes. The shock, no | Magnitude and timing of the demand arrival |
| Junk wax | Supply | No. Print runs were never published | Everything that mattered |
In two of three cases the mechanism was identifiable in advance and the trigger was not. That asymmetry is what a model can realistically be built to exploit.
Foil Research calculation from [1], [3], [4]. Summarises the three cases above.
The pattern across all three is that the mechanism was frequently knowable while the trigger was not. Grade convexity was visible in any population report. Supply asymmetry between modern and vintage was visible in any set's certified count. What was not visible was when demand would arrive, or how much of it.
A model built to forecast triggers will fail. One built to know which assets are structurally exposed when a trigger arrives, and how wide its own uncertainty is, is doing something achievable. That is the design constraint we work under, and we have published what it must clear before we would call it validated at the model page.
References
Sources.
- [1]
Vaulted Collection (2026). Charizard Pokémon card: prices, history and why it is so valuable.
Vaulted Collection
vaultedcollection.comReportingaccessed 2026-09-11↩ 1↩ 2↩ 3↩ 4
- [2]
- [3]
- [4]
All Vintage Cards (2024). The rise and fall of the junk wax era.
All Vintage Cards
allvintagecards.comReportingaccessed 2026-09-11↩ 1↩ 2↩ 3↩ 4
- [5]