What’s it worth?

Cards. Priced $25 or more, with at least 5 sales behind the new price (the old price came through the same screen; we keep the count behind the newest one), and at least 6 months of price history. Each move is the change between two medians of the same kind of sale — raw sold prices, same window — never a sold price against an asking price or a graded one.

We size every move two ways: a z-score against the card's own monthly moves over the last two years, and a z-score against every other card's move this week. The smaller one counts. Thin data shrinks it: the score is multiplied by √(n ÷ (n + 10)), so a move on 10 sales counts about 71% of the same move on many. A card makes the list when it is still more than one standard deviation outside its usual range; up to 5 each way, most unusual first, ties to the bigger %. Fewer than three qualify, we say so rather than pad. A card with no history cannot take a top slot; it shows only in a labelled thin-data tail when nothing qualifies.

Sets. The same two z-scores on each index's change this week, against its own monthly history since release and against every other set. Sets are never thin data, so the 5 biggest each way always show and a row inside its usual range says so. The dates on each row are the two price dates it compares.