Auctions don’t close cheap because they’re cheap. They close cheap because nobody qualified was watching.
SleeperLot scores auctions in the last minutes before they close: what the item is, what one costs to buy right now, and how far under that this lot is sitting. Low price on its own means nothing — forty bids is a correct price by definition. Thin bidding on an item we can independently value is the whole signal.
Most auction titles cannot be priced by anyone. The grey is the part of the day we refused to guess at.
Across — lots that closed under the cheapest credible Buy It Now ask, median —. That is what a buyer would have avoided paying by winning the auction instead of buying one outright — an asking price, not a sale.
Live from data/yesterdays-closes.json — the same file the
track record renders, generated from the corpus rather than typed
in. Last scored day: —
(—).
The declined column is the product
Most auction titles cannot be priced by anyone. No model number, no brand, nothing that pins the item to a comparable. We publish the decline rate on the front page because a tool that scores everything is guessing on most of its output, and a false positive here costs you money rather than attention. Rungs 4 and 5 of the matching ladder never reach an alert, whatever the apparent discount.
Ranges with a sample size, never a single number
You win the auctions where your estimate was highest, so guidance taken off a mean is
guidance to overpay. Every published figure is a range with its n attached,
and the figure we quote is the cheapest credible ask rather than the typical
one, because the conservative end is the only end that survives contact with the
winner’s curse. This is explained in full.
Every close published, including the ones we could not price
The track record is a uniform sample of a full day of closes, taken blind to price: hits, misses, and the ones we refused to score. It costs nothing to publish and it is the only way to know whether any of this works before you risk money on it.
Live alerts, when there are enough of them to be worth sending
The board shows what is closing under its reference right now. How often that fires is still being measured, and until the rate is known there is nothing worth charging for. Leave an email and you’ll hear once, when it is.