The record
Then I grade myself when it lands. Hits and misses, both published, neither edited afterwards.
Auction houses publish an estimate, sell the lot, and move on. Nobody ever goes back and checks whether the estimate was any good. There's no scorecard in this hobby. Nobody grades the grader.
So I do it. I call the price in public before the sale closes, show the comps I used, and post the result when the hammer drops. Sixty calls in, here's where that stands.
Beat the house
70%
Closer to the final price than the auction house was
A coin flip scores
15%
What a no-skill rule gets on the same lots. That's the number to beat.
Typical error
44%
The house runs 73%. So roughly 1.7× sharper than the people selling it.
And the part I'd rather not report: across 91 graded calls in July I landed 37% of them. Worse, the direction of my misses reversed mid-month — too low through the Finals sales, too high after. That's not a bias I can subtract. That's an overcorrection habit. The full month is here.
Called, published, waiting on a hammer.
Six calls. The bet is Siakam over Durant — evidence over name.
Twelve calls, made blind eighteen days out. Nobody had bid yet.
Twenty-four calls, all made before a single bid existed. The cleanest set I've done.
Sotheby's Christmas in July, 27 July. Four hits, two misses, one half. Both misses were high — and the worst of them is the one worth reading.
Best stat line in the sale. Finished last, at a third of Luka's price. I priced a box score like it was a moment. It isn't.
Dead centre. One below-tier comp, stepped up because that one was exhibition-worn and this was a real game jersey.
Game-worn jerseys closed at a median of $1,024. Game-issued — same team, same night, never actually played in — closed at $512. Exactly two times, across 26 lots.
The lots that run past everyone's expectations are attached to history: a Super Bowl-winning touchdown ball, a Finals-run jersey. A 39-point triple-double in an ordinary December game is a great night. The market knows the difference. I didn't.