Program Stats

A disposable on-chain program a trader deploys to run one strategy — almost always sniping new Pump.fun token launches — then closes minutes later to reclaim its rent, redeploying under a fresh id for the next run.

Why a program at all?

Sniping means "buy the instant the pool exists, atomically, or abort" — you can't do that reliably from a wallet. A tiny custom program bundles the whole attempt (and often multi-venue routing) into a single instruction that either lands complete or reverts.

Why thousands of failed transactions?

That's the race. The bot fires on every launch; most attempts lose the block or the token rugs, so they revert. The failures are the strategy — spray for the few that land.

Why redeploy and close?

The ~0.2 SOL of rent is refundable on close, and a fresh program id sidesteps any blocklist or reputation built against a known address. Cheaper and stealthier to burn identities than to keep one — so one operator can wear dozens of "new program" identities in a day.

How On Record catches it

Exact-bytecode dedup (same sha256 = same bot) collapses the redeploys into one cluster; lifecycle tracking sees the deploy → close; the failed-tx count confirms the intent. No explorer distinguishes "new protocol" from "same bot, 30th identity today" — that's a novelty-definition problem, which is exactly what this radar solves.