Do the finds actually go?

How this is measured, exactly

Every time the scanner runs, it takes a sample of the listings it flagged as underpriced and starts watching them. Alongside each one it also watches a control: a listing of the same card, from the same search, priced within 10% of the same fair-value estimate — one the scanner looked at and decided was priced about right. Both groups are then re-checked on an identical schedule, 1h, 6h, 24h, 72h after we first saw them.

What "gone" means, and what it does not. eBay's public API tells us whether a listing is still live. It does not tell us whether it sold, or for how much — that needs a separate eBay permission this account does not have. So a listing that disappears may have sold, or the seller may have pulled it. We are not going to dress that up as a sale: nothing on this page claims a find was bought by anyone, or names a price it changed hands at.

Which is the entire reason for the control group. "Most of our finds disappear quickly" would be a meaningless boast on its own — maybe every card listing does. The sellers-changed-their-mind problem, the listing-errors problem and the eBay-quirks problem all apply just as much to the control listings as to the flagged ones. So the number worth reading is the gap between the two, not either one by itself.

Where this is still soft. Times are upper bounds: we check at fixed intervals, so a listing recorded as "gone within 24 hours" may have gone in twenty minutes. The sample is a slice of each scan, not every find. And a fast disappearance is evidence that the market agreed the price was good — it is not a valuation, a guarantee, or advice to buy anything.