Monthly report · July 2026
Five sales, every call public before the hammer. Here's the month, including the part where I found out I'd been correcting the wrong way.
| Sale | Calls | In band | Median error | Misses ran |
|---|---|---|---|---|
| Goldin — June | 4 | 3 | −13% | Low |
| Sotheby's Finals, Part One | 5 | 2 | −71% | Low, all three |
| Sotheby's Finals, Part Two | 15 | 5 | −42% | Low, nine of ten |
| Sotheby's Summer Marquee | 64 | 23 | +42% | High, 49 of 60 |
| Sotheby's Christmas in July | 3 | 1 | +46% | High, both |
| July total | 91 | 34 | 37% |
Read that last column top to bottom.
Every miss in the first half of July was low. Every miss in the second half was high. The direction flipped completely, and it flipped in about ten days.
Here's what happened. I got beaten badly on the Finals sales — nine of ten misses too low, a Brunson jersey at 3.4 times my ceiling, a court panel at eight times it. So I adjusted. Then I adjusted too far, and spent the back half of the month calling everything high.
Which means the thing I published on 28 July — that I have a systematic high bias — was true of one sale and not of the month. I'd taken 60 lots from the Marquee and treated them as the record. The record is 91 lots, and it says something more uncomfortable: I don't have a stable bias. I have an overcorrection problem.
That's worse, because a stable bias is arithmetic. An overcorrection is a habit.
There's a second reading and I think it's the better one. Look at which sales ran which way.
The ones I called too low were all close to a live event — the Finals sales days after the Finals, Goldin at the end of the season. The ones I called too high were a broad vintage catalogue and a thin weekly in the dead of the off-season.
So it may not be a personal anchor at all. It may be that I'm not pricing when a sale happens. Material that's still warm from an event sells hot; the same category in a quiet week sells cool, and I've been applying one mental model to both.
That's testable, which is the only reason it's worth saying. Heritage closes on 23 August, in the pre-season dead zone. If the read is right, that one should run cool — and I've calibrated it downward accordingly. If it runs hot, the theory's wrong and I'll say so.
The market pays for the moment and discounts the name. Four of four six-figure vintage jerseys at the Marquee closed below their low estimate, including an Alcindor championship clincher. The two lots that blew past — a Super Bowl-winning touchdown ball and a Finals-run LeBron jersey — weren't the biggest names in the room. They were the biggest moments.
Worn beats issued, two to one. Twenty-six lots, same team, same night, ratio of exactly 2.00.
Three or more narrative hooks means the comp is a floor, not a ceiling. Learned expensively on a LeBron jersey at Goldin, and it's held every time since.
Sotheby's City and Classic closes 3 August. Goldin on the 8th. Grey Flannel on the 16th. Heritage on the 21st to 23rd, which is 24 calls made before a single bid existed — the cleanest set I've put out. And a charity sale on the 25th that doesn't count and I'm calling anyway.
Balls and strikes. Same as ever.