What the data taught us.
7 notes on file · occasional writing on how lectr reads the auction market — the methods, the measurements, and the wrong turns we kept the receipts for.
How we built the price-movement engine
A per-maker hedonic regression with a confidence gate bolted to the front — the controls, the abstention rules, and why only three makers currently clear the bar.
Read the noteHow lectr works
The systems walk-through — crawlers, the identity ledger, the engine, and the nightly replay, section by section.
Read the noteArt in Q2: trophies cleared, the middle got picky
A $509M quarter split almost evenly between Sotheby’s and Christie’s — the masterpieces cleared, about a third of everything offered failed to sell, and KAWS kept repricing.
Read the noteWatches in Q2: the deepest tape we track
1,811 watches sold for $240M at 97–98% sell-through — and the only vertical where the engine publishes a return: Cartier and Rolex up over five years, Patek down over three.
Read the noteDesign in Q2: small money, real heat
The smallest market we track posted the strongest demand of the quarter: 69% of sold lots beat their high estimate, Nakashima and Prouvé led, and the commodity end dragged.
Read the noteSports in Q2: the cards led a $126M quarter
A $2.93M LeBron rookie patch auto on top, a photo-matched Gretzky jersey behind it, 19,296 lots through the quarter — a card-led tape, read by volume and record.
Read the noteScience in Q2: between the seasons
Antique globes and Apollo Type I photographs led a cleaned tape — space and fossils ran hot on demand, meteorites soft, and the big seasonal sales land in Q3.
Read the note