Research · Mid-Summer 2026
In July I published that this market pays for the moment. In August I bet against my own rule eighteen times, and the market collected. The thesis held. I didn't — and that's the more useful result.
Six weeks of captured sales say the same thing three ways. A championship is worth 4.55× — ten bench-player jerseys from one game, one locker room, split only by which team won the title; the winners' median $6,400, the losers' $1,408, and the distributions don't touch. The same series' nets: the championship-clinching Game 5 net made $179,200, the Game 1 net $89,600. Exactly twice.
And the sharpest version, because it isolates the thing itself. Two Last Dance Air Jordan XIIIs, both worn, both signed. The pair only attributed to the playoff run, with no photo match, made $204,000. The pair conclusively matched — to an ordinary January game — made $84,000.
The match isn't what pays. The match to a moment is what pays. I wrote that on 30 July. Hold onto it.
Goldin's Summer Game Used sale closed 8 August. I called eighteen lots before the hammer. Seven hit, eleven missed — thirty-nine percent, all archived, none edited. The graded card is on The Record.
Three of those misses matter more than the rest, because they were the same mistake wearing three jerseys. I looked at a big number attached to a Curry pair and paid up for a moment that wasn't there.
| The lot | What I read into it | My call | Closed |
|---|---|---|---|
| Curry — 2016 Finals Game 1 | “The Finals.” | $46–68K | $28,500 |
| Curry — a 51-point night | “A milestone.” | $22–34K | $16,500 |
| Curry — first MVP season | “The MVP year.” | $6–12K | $5,500 |
Every one closed under my floor. Prices are hammer. On the face of it, my headline finding just took three straight hooks to the chin.
Read the middle column again. A Finals Game 1 is not the Finals — nobody lifts a trophy in Game 1. Fifty-one points is a box score, not an instant anyone remembers a year later. An MVP season is a hundred games averaged into an award; it has no single moment you can hold.
These aren't moments. They're stat lines. Impressive, real, worth something — and worth exactly what the market paid, which is the price of a very good game-used Curry pair with no iconic instant welded to it. The room read the letter, found a box score where I'd talked myself into a milestone, and priced the object. My own January-game Jordan — the conclusive match that made $84,000, less than the unmatched playoff pair — had already told me this would happen.
A stat line is not a moment. The market has never confused the two. I did, three times in one sale, with the rule sitting in my own published research.
So the thesis didn't crack at Goldin. It got its cleanest confirmation yet — at my expense, which is the only kind that teaches you anything. The moment commands a premium. The impostor does not. The entire skill is telling them apart before the number is up, and the misses are the map of where I still can't.
The rest of the sneaker misses ran the opposite way. Four ordinary game-used pairs — LeBron and Curry, no rarity, no moment, the kind of lot I'd called at four-to-eight grand — closed at $14,750, $15,000, $16,100 and $24,001. Two to three times over. Four for four.
There's no moment to misread here, because there isn't one. This is the floor of the game-used sneaker market lifting under its own weight — a broad, dull, upward repricing of the ordinary lot that has nothing to do with provenance and everything to do with demand outrunning supply in the cheapest tier. It is the exact inverse of the stat-line miss: there I paid for a story that wasn't there; here I ignored a tide that was.
I'm not publishing that as a finding. It's one sale, four lots, a hypothesis with a number on it, on the clock for the next auction to confirm or break. But it's named, not buried — because a rising floor and a mispriced moment are two different errors, and a valuation that can't tell them apart is worthless.
Two more misses were Michael Jordan, and both ran hot — a Dec 1995 pair I called at $70–95K that made $142,000, and his final-season Wizards jersey, which I'd talked myself down to $45–70K on a category discount and which closed at $150,000. Neither is a clean moment. Neither is the floor. It's the name doing its own thing — blue-chip Jordan pulling ahead of every model I put in front of it.
I could fold that into one of the two boxes above. I won't, because it doesn't fit, and the fastest way to ruin a dataset is to round the thing that doesn't fit into the thing that does. It's a third thread, tracked separately, waiting for more lots before it earns a sentence I'd stake anything on.
The insight isn't “the moment matters.” Everyone in this hobby feels that in their gut; it doesn't need me. The insight is that this market prices three different things and consistently confuses them at the edges — and that the confusion is where the money is.
| The market pays for… | …and shrugs at | Where I got caught |
|---|---|---|
| The moment — the clinch, the icon, the match to it | the stat line dressed as one | paid up for three Curry box scores |
| The floor, quietly, across the ordinary tier | nothing — it just rises | called four ordinary pairs far too low |
| The name, when it's blue-chip enough | the model you brought | discounted Jordan and got run over |
Three forces, three different errors, one sale. The job — the whole job — is naming which one you're looking at before the hammer falls. I'm keeping score in public precisely so you can watch me get better at it.
The dataset behind all of this now runs to 5,769 lots across eight houses, every one captured from the house's own published results page — the full sale, not the lots I happened to call. It is deliberately tiered and must be read that way: a full-population layer that can be pooled (Sotheby's, and the full runs at Heritage and Goldin), and a value-slice layer that only holds lots above a floor and so is a record-and-comp set, never a market size. I won't hand you a single headline dollar figure off it, because summing a truncated top-slice into a “market” number is the first lie these reports usually tell.
What it's strong enough to say: photo matching has now replicated as a premium across three separate years at Heritage — 450 lots, positive every year. That one's earned its place. The estimate finding — roughly two-thirds of everything with a published range sells below it — still rests on 219 lots from the only two sales that published ranges at all, and it stays flagged as such until a third house lets me test it. The Data page shows the shape of the whole set.
Goldin was eighteen calls. That's a scorecard, not a census, and two-thirds of it are the misses I'm building on — so treat every pattern here as a hypothesis with a next-sale test attached, not a law. The floor finding is one sale and four lots. The name thread is a handful. The sample still leans basketball, and live-floor rooms resist this whole method, because a visible bid stops meaning much once people are shouting in a room.
I'd rather say that here than have you say it for me. A record that only shows its hits isn't a record. This one shows the day I ignored my own rule and paid for it, because that day is the most valuable thing in the file.
The calls keep going up before the hammer, graded after. This is the mid-summer read; a full Q3 in Review follows in October, and the flagship — the one staked on the whole record — in January. Between now and then the only question that matters is the one above: moment, floor, or name. Balls and strikes. Same as ever.