The launches that hurt the most are decided inside a single block. The token deploys, and in that same block, often in the same transaction bundle, a cluster of wallets buys a large slice of the supply at the opening price. By the time the pair shows up on a screener and you are looking at a chart, those wallets are sitting on a quarter of the float at a cost basis nobody else will ever touch. Everything that follows, the posts, the volume, the plateau that looks like accumulation, is distribution. Retail demand walks in and the bundle walks out into it.
The part that keeps me interested is that this setup leaves fingerprints all over the chain, and reading them takes a few minutes once you know where to look. The deployer wallet is the thread you pull first.
What a bundle looks like on-chain
The mechanics vary by chain but the shape is consistent. On Solana, an operator funds a set of fresh wallets, deploys the token, and submits everything atomically so the sniper buys land in the same slot as the deployment, ahead of any public transaction. On Ethereum the same trick runs through block builders. Either way the signature is identical: a group of wallets that bought in the very first block, at launch price, before the token was visible anywhere.
Individually these wallets look harmless, which is by design. A naive top-holders check shows thirty or forty wallets each holding one or two percent of supply, and that passes most people's sniff test. The question that actually matters is whether those forty wallets are one entity wearing forty coats, and funding history is how you answer it.
Tracing the funding chain
Take every wallet that bought in the launch block and walk its funding backwards. Where did the SOL or ETH that paid for the buy come from, and where did that sender get its money before that. You are looking for convergence on a common parent within one or two hops.
Lazy operations fund the snipers straight from the deployer. You will see the deployer receive a lump sum, fan it out in near-identical amounts to a dozen brand-new wallets in the minutes before launch, then deploy the token. This is trivial to catch on any block explorer, and it is still remarkably common, because most buyers never look.
More careful operations insert an intermediate wallet or two between the deployer and the snipers. The chain still converges if you keep walking. A funding wallet that fans out to fifteen fresh addresses which all buy one token in its launch block and then do nothing else is one operator, whatever the hop count says.
The hard case is funding routed through an exchange. A withdrawal from a CEX hot wallet breaks the on-chain link, since everyone's withdrawals come from the same pool. Even then there are tells. Near-identical withdrawal amounts landing in fresh wallets within a tight window, all of which then sit idle until the exact launch block of the same token, does not look like organic demand. Timing and amount correlation is weaker evidence than a direct funding edge, and I weight it accordingly, but week-old wallets whose only life event is this one launch have never once, in my experience, turned out to be forty separate enthusiasts.
The deployer's prior launches
The second half of the workflow is the deployer's own history, which is usually more damning than the funding graph. Pull every token that address has deployed. Serial deployers are easy to spot: a new token every few days, each one with the same vertical open, the same brief plateau, the same bleed toward zero within a week. That is a production line, and the current launch is the next unit off it.
Also check who funded the deployer itself. If the deployer's gas came from a wallet that also funded five other deployers, you have found a farm, and every launch out of it inherits the track record of all the others. A fresh deployer with no history is genuinely ambiguous, and I want to be honest about that. No rap sheet means no evidence either way. Some legitimate teams also bundle deliberately, buying their own launch block so third-party snipers cannot take the supply, then locking or slowly redistributing it. So a bundle on its own is a yellow flag rather than a verdict. What separates the two cases is what the bundled wallets do afterward. Holding, burning, or sending to a locker reads very differently from spraying tokens across fresh wallets that begin selling in coordinated clips.
The five-minute check before entry
- Pull the launch block and list every buyer in it. Add up their combined share of supply. If same-block buyers hold more than roughly 15 to 20 percent of the float, treat the launch as bundled until proven otherwise.
- Walk each launch-block buyer's funding back two hops. Any convergence on the deployer or a shared parent collapses those wallets into one holder, so redo the concentration math with that in mind.
- Check wallet ages and funding amounts. Fresh wallets funded with near-identical sums minutes before launch are one operator.
- Pull the deployer's deployment history and the outcomes of its prior launches, then check who funded the deployer itself.
- Look at what the bundled cluster has done since launch. If it is already distributing, you are the exit liquidity it was waiting for. If it is untouched, you at least know the overhang exists and can size for it.
The most common failure mode I see is people running none of this because the chart is moving and they are afraid of missing the entry. Which is exactly the state of mind the bundle is priced for. The plateau after the first leg up exists to give you time to talk yourself in.
None of this catches everything. Operators adapt, funding graphs get deeper, and some rugs come from tokens with clean launches and dirty liquidity instead. But same-block concentration plus a converging funding graph plus a deployer with a body count filters the large majority of launch-day traps, and it does so before you have any money at risk. I built cluster views into Blockcircle's wallet tracking partly because I got tired of doing hop-by-hop walks in an explorer tab, but the manual version works fine and costs nothing. Five minutes reading the deployer usually tells you more than an hour reading the chart.