Forty-eight hours after the archive got eyes, the first patterns

Yesterday I wrote that the launch archive had started capturing holders, buyers, and transaction velocity, and that I didn’t know yet whether any of it separates survivors from corpses. The week-mark verdicts that settle that question start landing Monday. But two days of feature data already shows some shapes worth writing down.

The clearest one: crashed launches move more transactions through fewer wallets. Among the six-hour snapshots sampled so far, launches that crashed inside their first twelve hours averaged 824 transactions with a 0.41 unique-buyer ratio in the sample — meaning fewer than half the sampled trades came from distinct wallets. The ones that hadn’t crashed averaged 501 transactions at 0.62. High activity from few actors is what wash trading looks like, and it shows up before the crash flag does. About one launch in five samples below 0.3 — the same handful of wallets cycling a token between themselves to look alive.

The starkest single row: one six-hour-old token showed 7,665 transactions, 1,134 holders, and effectively nothing left in the pool. Over a thousand people holding a token with no liquidity behind it, at hour six of its existence. Three of the five busiest six-hour snapshots in the archive share that shape — enormous velocity with nothing underneath by the time the snapshot fires. The other two carried real pools, which is the point of measuring: velocity alone doesn’t tell you which one you’re looking at.

Also now measured instead of assumed: comebacks are rare. Of 695 launches that crashed hard in their first half-day, exactly 2 ever traded at three times their six-hour price afterward. The early crash has been, so far, close to a one-way door.

Two housekeeping notes from the same 48 hours. The discovery screen’s liquidity floor moved from $1,000 to $2,000 — the first night’s data showed 41% of passing launches sat in that band, spending instrumentation budget on pools too thin to matter; the threshold moved on that receipt, forward-only, with every launch’s row recording the values it was judged against. And the archive page now carries a “pre-watch” roster: launches that are actually holding up — a day old or more, a real pool, price holding near its peak or recovered off an early crash — surface automatically as the data accrues. At birth the roster was honestly empty; the first launches old enough to qualify age in about now.

All the usual caveats at full strength: two days of data, no verdicts yet, sampled ratios, and correlations found by looking. These are leads, and the archive will get to test them against real outcomes starting next week. That’s the whole reason it exists.

— Adam, Studio Amadeus