Methodology

1. Where the data comes from

We read the open card-data sources that aggregate the major marketplaces: TCGdex (Pokémon), Scryfall (Magic), YGOPRODeck (Yu-Gi-Oh!), Lorcast (Disney Lorcana), and TCGCSV — TCGplayer's daily catalogue — for One Piece, Digimon, Star Wars Unlimited, Dragon Ball, Gundam, Sorcery and Flesh and Blood. Between them these carry TCGplayer and Cardmarket pricing. Prices are fetched live on each lookup; the snapshot-backed games use a daily file whose date is printed on the quote.

2. The consensus fair value

We never trust a single number. For each card we gather every available price signal — market, mid and direct-low from TCGplayer; trend and 7/30-day averages from Cardmarket; equivalents from the other feeds — and convert everything to USD.

Those signals are then blended by an outlier-resistant weighted consensus. We first compute a weighted median as a robust centre, discard any signal more than 2.5× above or 0.4× below it, and take the weighted mean of what survives. That discard step is the important one: a single mis-keyed listing at eight times the real price cannot move the result. The surviving figure is the fair market value, quoted for a Near Mint, ungraded copy.

3. Confidence, and why it is often not 100%

Every value carries a 0–100% confidence score built from two things: breadth (how many distinct sources agree) and agreement (how tightly they cluster). Discarded outliers reduce it. Crucially, a lone price signal is capped well below a corroborated one — a single number cannot corroborate itself, and zero measured dispersion there means "nothing to disagree with", not "perfect agreement". When confidence is low we round the headline figure and show a range rather than implying false precision.

4. Condition adjustments

Raw condition multipliers are applied relative to Near Mint, and they are era-aware, because a played vintage card holds far more of its value proportionally than a plentiful modern one:

Era	LP	MP	HP	Damaged
Vintage (≤2003)	0.90	0.78	0.62	0.45
Retro (2004–2010)	0.87	0.74	0.58	0.42
Modern (2011–2019)	0.85	0.70	0.55	0.40
Current (2020+)	0.82	0.65	0.48	0.35

5. Graded values — a model, not a comp

This is the part to read most carefully. We do not have per-grade sold comps or population data. Graded figures are modelled from the raw market price using three factors:

• Condition value — what the card is worth on its condition alone at that grade.
• Premium share — how much of the gem-mint premium a given grade captures. A 10 captures all of it; a 9 captures a small fraction.
• Era gem premium — how much a flawless copy commands over raw in that era. We use roughly 8× for vintage, 4.2× retro, 2.6× modern and 2.1× current, because vintage gems are genuinely scarce while modern print quality makes high grades common.

That result is then scaled by the grading company. PSA anchors the scale on both liquidity and premium; BGS sits just below; CGC and SGC are modelled a step lower on premium while remaining liquid; budget graders are discounted sharply, because in practice they add protection rather than resale value. A floor ensures a gem slab is never valued below the raw card plus the cost of grading it.

These weights are estimates inferred from market behaviour — not published facts. That is precisely why graded quotes carry a reduced confidence score and are labelled as estimates.

6. Buy and sell prices

Fair value is what a card is worth; it is not what you should pay. We model the realistic bid–ask spread a disciplined trader captures and place the buy price below fair value and the sell price above it. The spread widens with illiquidity and with measured volatility, and momentum nudges both sides. Sell figures are shown net of marketplace economics — roughly 13% in fees plus a fixed charge, plus value-scaled shipping — because gross price is not what you keep.

7. Momentum and volatility

We log a price point for every tracked card daily and build our own history. Momentum compares a robust recent level against a longer average; volatility is the dispersion of that series. Before either is computed the series is outlier-cleaned against its own median, and implausible swings are discarded as data errors rather than reported as moves. Where a card has too little history, we say nothing instead of guessing.

8. What we deliberately don't do

• No realised auction history. The free data we use is price-guide based, so values reflect current market consensus rather than individual recent sales. Sites with paid sold-comp feeds will differ, particularly on graded cards.
• No population data, so we publish no market caps or indices — those require scarcity data we don't have.
• No inflated numbers. Affiliate links are generated after a value is computed and never influence it, and we link to competing marketplaces.
• No invented precision. Where a card is thinly covered we widen the range, lower the confidence, and say so.

9. In plain terms

SplitShot is an estimate engine for guidance, not an appraisal service. Values move, sources lag, and individual sales vary. The honest use of this site is as a reference point to check an asking price against — which is exactly why we show you the confidence, the range, and the sources behind every number.

Value a card →