How we price.

The sources

Every valuation on GrailSpy is drawn from these signals, in this priority order:

1. Sold history — actual eBay AU sales. Confirmed completed sales with real sold-at timestamps. Median over a trailing window (7-, 30-, or 90-day, depending on the surface). What a card has actually sold for, not what sellers are asking.
2. Active listings, identity-matched. Used as provisional comps when no in-window sold history exists. Same set, same number, same finish. Asking prices, not sales — surfaced because thin data still beats no data, and clearly flagged as provisional.
3. Title-match fallback. Looser match across recent active listings when nothing above is available. Broadest coverage, weakest signal. Also flagged as provisional.

What we adjust for

• Condition. Four buckets — NM / LP / MP / HP — parsed from listing titles. NM is the baseline. When a non-NM condition has fewer than three sold comps in the window but NM is direct, we extrapolate using fixed industry-heuristic ratios (LP ×0.85, MP ×0.70, HP ×0.50) and label the slice as extrapolated. These ratios are placeholders — they will be retired once we have enough AU sold data to learn the real curve.
• Currency. AU pricing only. We filter to AUD-denominated sales and listings; we do not currently convert from USD or JPY. Cross-currency comparisons are not in alpha.

What we surface

• Median price — the middle of recent sales. Less skewed by outliers than the average. Trimmed: sales outside [0.5×, 2×] of the initial median are dropped before the final median is taken.
• Sample size — the count of in-window parsed comps for the primary condition. Always shown. Drives the confidence flag.
• Confidence — a limited-data badge fires when the primary condition has fewer than five in-window comps. Anything at or above five is unflagged. We do not surface a "high confidence" label — alpha samples are too small to earn one.
• Trend — when both a 7-day and a 30-day window have data for the same condition, the ratio is labelled rising / falling / stable with a ±5% deadband around 1.

What we don't do

• We don't extrapolate aggressively. If the data is thin, we say so on the row.
• We don't quietly mix asking prices with sales. When we fall back to active listings, we mark the result provisional.
• We don't conflate raw and graded cards. Alpha pricing covers raw NM/LP/MP/HP only — PSA / BGS-graded valuations are not yet in scope.
• We don't forecast. We report what happened in the window and flag the shape of it. You decide what happens next.

The honest admission

The AU TCG market is small and noisy. Sample sizes are limited. Our confidence flagging is conservative on purpose. When a card has no in-window sold history we either fall back to provisional active listings or render the value as unavailable — never as a fabricated zero.

Questions?

Email privacy@grailspy.com.au. We'll tell you what we know and what we don't.