How this was measured

Everything below is stated so the numbers can be argued with. If a step looks wrong to you, the per-launch rows are on /v1/study/launches — recompute it.

The question

When several wallets from a "smart money" set buy the same new pump.fun token within seconds, does that token graduate more often than it otherwise would?

Sample

4,000 launches drawn from a 21-day window.
Outcome measured: graduation (the token completing its bonding curve).
Observed base rate across the sample: 3.3%.

The tracked set

A wallet enters it by trading history only — no manual picks, no follower counts: at least 15 distinct tokens traded, at least 8 closed round-trips, and a median realised return above +3%. That is a deliberately low bar, which matters: it means the set is large, and a large set makes accidental co-occurrence more likely. Hence the control.

The control group

The comparison is not "smart wallets vs nothing". It is against random active wallets that are not in the set, sampled to match the tracked set's share of the early-buyer stream. So both columns face the same token, the same moment, and the same crowd size — only membership differs.

Definitions

early_buyers — distinct buyers in the first 60 seconds.
smart_peak30 — the most tracked wallets seen inside any 30-second window.
placebo_peak30 — the same measure on the control set.
Quiet zone — fewer than 30 early buyers. Busy zone — 30 or more.
Intervals are Wilson 95%.

What we found, including the part that fails

Quiet zone: convergence separates from both baseline and placebo, and rises with each additional wallet.
Busy zone: it does not work. Tracked wallets score the same as random ones. On a busy token "smart money bought it" mostly means "a lot of people bought it".

Limitations we would rather state than bury. Single 21-day window, so regime change is not ruled out. Above threshold 3 the placebo group has few launches left, so those rows show "no rise observed" rather than a precise zero. Graduation is not profit: it says nothing about entry price, slippage, or what the token did afterwards. Nothing here is advice, and none of it is a claim that you can trade it profitably — that is a separate question with a separate answer.

Reproduce it

curl https://api.smugcalls.com/v1/study/launches?limit=25 returns the rows the aggregates are built from: buyers, tracked count, placebo count, outcome. Group them however you like. If your numbers disagree with ours, we would like to know.

← back to the numbers