How a verdict is made.
Seven signals are scored and blended into a composite. A regime classifier then decides what kind of market the card is in, and the regime, not the composite, sets the verdict and the price band. This page states those rules in plain English, without publishing the exact dial settings.
Reprint Trough vs Structural Decline
Two cards can have the same falling chart and opposite correct answers. The whole tool exists to tell them apart. A falling price normally reads as a sell; in a reprint trough it is usually a hold, because reprint supply washes through and popular cards recover.
Reprint Trough (popular)
Price is falling because supply rose (a reprint or restock), not because demand died. The price band sits above today's price, because a recovery is expected.
All four must hold. If the set's popularity is weak, this is Structural Decline wearing a disguise: the popularity floor is what separates them.
Structural Decline
Price is falling and no supply event explains it: demand itself is fading, with no scheduled end to the pressure. The price band sits at or below today's price.
A confirmed reprint on the same dates would reclassify it, which is why reprint evidence is reviewed rather than scraped and trusted.
What "falling" means, precisely
A downtrend needs both horizons to agree. The short-term slope alone used to be the only gate, and a short wobble could overrule everything else: a card dipping over a short window while trending up over a longer one was called a Structural Decline. A dip inside a rising longer-term trend is a pullback, which is a different thing and gets its own path.
A pullback resolves to Accumulation when reprint supply is exhausted and scarcity clears its floor, and to Stable / Illiquid otherwise. It is never a Structural Decline.
The seven weighted signals
Each is scored on a common scale, where higher always means a stronger case for holding. The composite is a straight weighted sum. Reprint pressure and popularity deliberately dominate the blend, because between them they decide the fork above.
Composite, and the holding/selling gates
Both sides are built explicitly and the disagreement is named. Signals that back the verdict are printed first because they explain it; then the single strongest signal that contradicts it is printed as a labelled counterpoint. When nothing at all supports the call, the card says so and is tagged trend only: a verdict resting on price trend alone is a much weaker claim than one with signals behind it, and you are entitled to know which you are looking at.
Six regimes, six verdicts, six bands
The regime is decided first and it drives the verdict. The composite only breaks ties inside Stable / Illiquid, overrides a decline when fundamentals are exceptional, and decides Rotation Fade.
| Regime | Verdict | Price band | Trigger |
|---|---|---|---|
| Spike / Distribution | Sell Now | above today's price, expecting the spike to partly hold | a sharp short-term price spike, with enough recent sales to trust it |
| Reprint Trough (popular) | Hold Long | above today's price, expecting a recovery | falling, reprint active or recent, popularity clears its floor, set historically bounces |
| Structural Decline | Sell Now | at or below today's price | falling with no reprint to explain it, unless the composite is unusually strong, which downgrades it to Hold Short |
| Accumulation | Hold Long | above today's price | reprint supply exhausted, scarcity and popularity both clear their floors |
| Rotation Fade | Hold Short or Sell Now | roughly today's price, tilted slightly below | rotation pressure is high; sells only if the composite is also weak |
| Stable / Illiquid | Hold Short by default | a wide band around today's price | no dominant trend, or too few recent sales to trust one; a strong composite tilts it to Hold Long, a weak one to Sell Now |
Bands are percentages of the card's current market ask, and they are regime-dependent on purpose: a trough is expected to recover above today's price, a decline is expected to trade below it.
Every gate, named
Why the override sits high. A genuine two-horizon decline is the model's clearest sell signal, so only an unusually strong composite should override it. A lower bar caught ordinary declines and made Sell Now nearly extinct. A threshold that silently deletes a verdict is worse than no threshold.
Confidence and liquidity
Confidence is not a feeling about the verdict. It is a count of how much real evidence exists behind it.
The confidence ladder
These are the terms that add confidence. Evidence that is present but degraded (most obviously a price quote old enough that it no longer describes the current market) is discounted rather than counted, so the grade a card carries can never rise as its evidence gets staler.
Liquidity flag
- Healthy: regular trade in sold comps; the price means something.
- Thin: few sold comps; price uncertain.
- Illiquid: almost no sold comps. Treat the price as indicative only.
Thin markets make any single price a wide estimate. On a Stable / Illiquid card, read the band, not the midpoint.
What you paid is a fact about you, not the market
When you record a cost, your unrealised gain is reported alongside the verdict, but it is deliberately walled off from the signals, which know nothing about your entry. A card can be a perfectly good hold and still be worth trimming after a big enough run-up.
- Up substantially: a note appears suggesting you consider taking some profit, even on a hold.
- Up a great deal on a Hold Long: the verdict shortens to Hold Short so you revisit sooner. It is never forced to a sell: "it went up a lot" is not by itself a reason to sell.
- Down sharply: a note appears stating plainly that the signals describe the market, not your entry.
Every label the app shows
Read live from the desk's own glossary, so a definition here can never drift from the code that produces the label.
The questions that come up.
Why is a falling card ever a Hold?
Because falling price has two causes and only one of them is bad news. A reprint or restock raises supply, which pushes price down for a known period and then stops. Demand leaving is different: nothing ends it on a schedule. The classifier requires an actual reprint state, a popularity score that clears its floor, and a set with a documented history of recovering before it will call a falling chart a Hold Long.
Why does a Sell Now sometimes print alongside signals that say hold?
Because those are two different systems and they are allowed to disagree in public. The verdict comes from the regime, which is a rule about price trend. The signals describe fundamentals. When they conflict, the strongest opposing signal is printed as a named counterpoint rather than dropped, and the verdict is tagged trend + signals or trend only. A call the fundamentals do not support is still shown to you, labelled as what it is.
Does "Sell Now" mean sell today?
No. It means conditions argue for selling into current demand. It is directional, not urgent, and it is a research aid rather than advice. Hold Short means hold but revisit within weeks, usually an unresolved catalyst such as a reprint washing through or a spike that may retrace. Hold Long means hold through the current cycle.
Where do the prices come from?
A local archive of weekly TCGplayer prices going back to 8 February 2024: millions of real observed points, queried without a network call. It is the pipeline's primary trend source. Because it is TCGplayer, it is an English-market archive: a Japanese restock is real news our data can neither confirm nor deny, and the app says so instead of guessing.
What happens when a card has no real price history?
A plausible curve is generated around the real current price so the page has something to draw, and the classifier is forbidden from reading it. Directional regimes are gated behind a real series. Cards in that state fall through to the fundamentals-driven regimes and carry lower confidence, which is exactly what the confidence ladder is for.
How is popularity decided?
Durable collector demand for the set and the Pokémon, curated by hand for a minority of sets; every other set uses a neutral default. That matters, because the popularity floor is what separates a Reprint Trough from a Structural Decline; a set on the default cannot clear it alone.
Are reprint signals trusted automatically?
No. Automated feeds can only ever file a signal as pending. Nothing scores as fact until it is confirmed by review, signals can be marked rumor or rejected, and every decision is reversible. A regime that hinges on a reprint should not hinge on a scraper.
Why does the Last Sold layer keep coming back empty?
A true completed-sales median needs eBay's Marketplace Insights API, a restricted scope these credentials have not been granted; the request returns invalid_scope. The layer is reported as unavailable with that reason attached and fills in automatically if access is granted. It is never backfilled from asking prices, and the ask/sold spread stays switched off rather than measuring asking against asking, which measures nothing.
Does the graded price mean a slab sold for that?
No, and it is labelled everywhere it appears. Graded numbers are asking prices on active eBay listings, trimmed for outliers, and graded slabs list well above what they clear. They are never allowed to stand in for a sale.
Is grading worth it?
Usually not, and the model says so. Expected value is blended across grade outcomes and charged the real PSA fee, the real tier, shipping both ways and the turnaround. A card only counts as a grading candidate if the net upside beats a meaningful share of its raw value. Assuming a cheaper fee tier than you can actually order makes grading look viable on cards where it is not.
Does this work for games other than Pokémon?
Catalogue and price data, yes. Verdicts, no. The reprint brain, rotation logic and era rules were fitted to Pokémon, and running them against another game would produce confident output from rules that were never calibrated for it. The card page states that plainly rather than letting the interface imply otherwise.
Is this financial advice?
No. It produces directional signals and estimates for trading cards, it will be wrong sometimes by design, and you make your own decisions. See the Terms of Service.