How it works
About a minute after a Pump.fun token is created, SNIPERWATCH reads its first 120 seconds of trades from the chain, traces where each early buyer’s SOL came from, and scores how much of the supply sits with wallets that look coordinated. Ten minutes in, it scans again to catch late bundles. Both results are kept.
Signals
- Launch block
- Bought in the creation slot or within 2 slots. Humans clicking a fresh listing rarely land there; scripts do.
- Common funder
- Two or more buyers received SOL from the same wallet in the 24 hours before buying, directly or two hops back. A shared fee payer counts too. Wallets are grouped into clusters with union-find.
- Dev-linked
- The buyer received SOL from the creator, sent SOL to the creator, or shares a funder with the creator.
- Fresh wallet
- The wallet’s first transaction was within 24 hours of the buy. On its own this never flags a wallet.
- Uniform size
- Three or more buyers spent nearly the same SOL (within 2%). Bundlers like round, repeated numbers.
- Repeat
- The wallet was flagged on earlier launches. Three or more earns the Repeat Offender badge.
- Funding origin
- Deposits from bridges (Relay, Wormhole, deBridge, Mayan) and exchanges are labeled, with a link to the bridge explorer that shows the source chain. They never link wallets on their own.
- Serial launcher
- The creator was funded by a wallet that also funded the creators of other launches. Counts toward the repeat component.
Snipe Score
Each component scales linearly from zero to its weight and caps at the saturation point. The score is their sum, from 0 (clean) to 100 (fully bundled). Launch farms get a floor: when the creator was funded by a wallet that also funded other launches’ creators, the score is at least 25 + 10 per other launch, up to 70. Wallets ignored as infrastructure: exchange hot wallets, Pump.fun program and fee accounts, Jito tip accounts, and any account sending more than ~30 transactions a minute (bot relayers, unlisted exchanges).
| Component | Weight | Maxes out at |
|---|---|---|
| Supply held by linked clusters | 40 | 40% of supply |
| Supply bought in the launch block | 20 | 30% of supply |
| Supply held by dev-linked wallets | 20 | 20% of supply |
| Share of fresh wallets among buyers | 10 | 60% of buyers |
| Repeat offenders present | 10 | 3 wallets |
0–20 CLEAN21–45 CAUTION46–70 BUNDLED71–100 COOKED
Limits
- False positives happen. Friends funding each other from one wallet look like a cluster. Exchanges we have not listed can look like a shared funder.
- Evasion works. Funding through many hops, privacy tools, or long-aged wallets defeats a 24-hour, two-hop lookback.
- Holdings are net early positions: buys minus sells inside the window, not live balances.
- Very busy wallets (thousands of recent transactions) are treated as old, and their funding may be missed.
- A score describes wallet behavior in public data. It says nothing about who controls a wallet or why.
Report a false positive
Wallet owners can flag a wrong result. Reports are stored and reviewed against the evidence on the scan page.