what happens to a token when a fomo whale buys it first
a first buy from a large fomo account is not a normal market order. we timed 29 of them to the minute. here is the curve, the test that separates skill from cascade, and everything that is wrong with the sample.
on the 24th of august 2026 we pulled every first buy made by one wallet near the top of the fomo leaderboard over the preceding month, and timed what the token did afterwards. a first buy means the first time that wallet ever bought that token. not an add, not a re-entry. the first one.
we did it because we wanted to answer a specific question, and the answer turned out to be more interesting than the question. we wanted to know how fast a copy trade has to be. what we found was a decent argument that the thing being copied is not what most people think it is.
the finding
read left to right, that is a move that is most of the way in by the first minute, peaks around a quarter of an hour, and has given a large share of itself back by the end of the hour. the last figure is the one people skim past, so here it is in words: fifteen minutes after the fill, the price was sitting at the 93rd percentile of everything it would print that hour. sixty minutes after the fill, it was sitting at the 57th. near the top, then barely above the middle.
for anyone trying to copy these trades by hand, that curve is the entire problem. a thirty second delay costs you a slice. a thirty minute delay puts you on the wrong side of the peak. eight hours of sleep puts you nowhere near the trade at all.
how we measured it
the method is deliberately boring, because interesting methods are how you talk yourself into a result.
- subject. one wallet,
0x8224c04a…, resolved from its public fomo profile by matching displayed token quantities against on-chain holder balances. two independent quantity matches were required before we accepted the address as that trader's. - events. every first purchase of a previously unheld token in the window. 29 of them.
- clock. block timestamp of the fill, not the timestamp of any app-side event. the fill is the only moment that is not ambiguous.
- price series. on-chain trade prints for the same pool the entry filled in, sampled at +1, +15 and +60 minutes.
- statistic. the median, not the mean. one 900% outlier will drag a mean anywhere you like, and in a sample of 29 you should assume there is one.
the test that separates skill from cascade
a +43% median move has two competing explanations, and they have opposite implications for anybody thinking about copying.
explanation one: alpha. the trader is genuinely early. they find tokens before the market does, the market catches up, and the move is the market being right slightly after they were. under this reading, the trader's own buying is a small part of the move and their judgment is the asset.
explanation two: cascade. the trader's order is a starting pistol. it fires a push notification at a very large number of people, a meaningful fraction of whom buy within minutes, and the move is mostly them. under this reading the trader's judgment is close to irrelevant and their audience is the asset.
these two stories make a different prediction about one measurable quantity: the share of the minute's volume the trader's own order represents. if the move is the trader's own price impact, more of their volume should mean more of a move. if the move is the crowd, the cases where the crowd shows up hardest are the cases where the trader is the smallest slice of it.
| trader's share of that minute's volume | median move |
|---|---|
| under 40% — the crowd supplied the rest | +53% |
| 40% or more — the trader was most of the tape | +31% |
the inversion is the whole result. it is not proof — a small sample can invert for boring reasons, and thin tokens have their own pathologies — but it is the sign you would expect if the notification, not the order, is doing the work. the volume is the followers.
if that is right, then the leaderboard column everyone sorts by is the wrong one. realised pnl describes what a trader did. follower count describes how hard the next notification will land. a trader with half a million followers and unremarkable returns moves a token harder than a trader with forty thousand followers and excellent ones.
what the decay means in practice
put the two results together and the shape of the opportunity is unusually specific:
- the move is front-loaded. most of it exists by the first minute. arriving after the alert has already been read by everyone else means arriving into the second half.
- the move peaks early, around fifteen minutes, and then mean-reverts hard inside the same hour.
- the race is therefore not against the market. it is against the notification pipeline — push, phone, tap, confirm. anything that detects the fill on-chain, at the block, is ahead of every human in that queue by construction.
this also reframes what a copy tool is for. it is not a way to borrow somebody's judgment. it is a way to be at the front of a queue whose ordering is determined by reaction time, and to leave it on a rule rather than on a feeling.
what this does not prove
we would rather say this ourselves than have it said back to us.
- n = 29. that is one wallet, one month. the confidence interval around a median of 29 skewed observations is wide enough to drive a truck through.
- one trader. everything here could be an artefact of one person's particular style, size or taste in tokens.
- one regime. a single month of one market. the same measurement in a quiet tape could come back flat.
- survivorship at the top. we measured a wallet that is near the top of a public pnl leaderboard, which is a sample selected on the outcome. the leaderboard does not show you the accounts that ran the same strategy into the ground.
- the volume split is a correlation. low trader-share minutes might differ from high trader-share minutes in ways we did not control for, starting with liquidity depth.
what would change our mind
a thesis that cannot lose is not a thesis. these are the results that would sink this one:
- post-trade returns across many wallets show no relationship to follower count. the cascade story dies immediately.
- the volume split reverses at scale — bigger trader share, bigger move. that is the price-impact story, and it means the crowd is not the driver.
- the decay disappears in a larger sample. if the move holds past the hour, this is ordinary early-discovery alpha and speed matters far less than we think.
- first buys stop outperforming adds. the notification-as-news mechanism would lose its footing.
what we actually do with it
three things, and we would rather be explicit than vague.
we detect on-chain, not in-app. a websocket subscription to the chain sees the fill in the block. a notification pipeline sees it after push, delivery, unlock and tap. the gap between those two is where this entire measurement lives.
we copy the exit too. a curve that peaks at fifteen minutes and decays through the hour is a curve where the sell is worth more than the buy. a tool that mirrors entries and leaves you to find your own way out is selling you the losing half.
we print the slippage. your fill will be worse than theirs. on a violent candle, meaningfully worse. the number goes on the trade instead of into a dashboard nobody opens, because the alternative is a product whose only honest description is that it hides its own cost.
the surrounding argument — what fomo is, why the notification bus is the real engine, and why the leaderboard's follower column outranks its profit column — is laid out in fomo trading, explained. the practical filters we apply when picking a trader are in which fomo traders are worth copying.
frequently asked
how many trades is this based on?
29 first buys from a single wallet over roughly one month, ending 2026-08-24. that is a small, non-random sample from one trader, and it is the first thing to hold against every number on this page. it is enough to state a hypothesis and to justify a larger study. it is not enough to justify a strategy.
what counts as a first buy?
the first purchase of a token the wallet has never previously held. adds to an existing position are excluded, and so are re-entries into tokens the wallet had held and sold before. the distinction matters because the notification a first buy generates reads as news, while an add does not.
why would a smaller order produce a bigger move?
because the move is not the order. if the price impact came from the trader's own buying, a larger share of the minute's volume would produce a larger move. we measured the opposite. the most economical explanation is that the trader's order is the trigger and their followers are the volume, so the cases where the trader supplied little of the volume are the cases where the crowd supplied a lot.
does this mean copying is profitable?
it means the timing window is narrow and the entry price is everything. a median +43% at fifteen minutes that decays inside the hour is a great trade at second one, an ordinary one at minute ten and a bad one at minute forty. the measurement says nothing about whether you will be on the right side of that. it says where the clock is.
one second, not thirty. and the exit too.
if the curve tops out at fifteen minutes, the gap between their fill and yours is the whole trade. copyfomo closes it to about a second, both directions.
open copyfomo on telegram →