Detecting wallet operator changes
A profitable public wallet can attract copytraders. If control of that wallet changes, the address stays the same while the transaction builder often does not. The indexer makes those construction details queryable, so monitoring can look beyond PnL and token selection.
This experiment compares two adjacent 24-hour windows for active Pump.fun wallets. Its fingerprint uses tip routing and size, compute-unit behavior, priority fees, instruction ordering, fee-payer behavior, instruction variants, and parent-program mix. Each wallet needs at least 20 transactions per window. The examples float with live data and are regenerated by the documentation build.
Window end: 2026-08-23T11:40:41Z · Comparable wallets: 250
Current tiers
Thresholds are calibrated conservatively across the current population. Routine variation stays stable, ambiguous or internally noisy profiles go to review, and only the extreme high-change tail enters the changed tier.
| Stable | Review | Changed |
|---|---|---|
| 231 (92.4%) | 16 (6.4%) | 3 (1.2%) |
High-change tier examples
These wallets are investigation candidates, not claims of a wallet sale or bad behavior. A bot upgrade, routing change, RPC change, or strategy change can create the same signal.
| Wallet | Transactions, old → recent | Fingerprint, old → recent | Change score | Within-window stability | Main changes |
|---|---|---|---|---|---|
7h2TDC…yGHNE9 | 744 → 777 | fc52d75fcbec → 2751aeae1144 | 37.1 | 71.7% | average tip, instruction order, preferred tip route |
9GL9rj…EaeE2L | 1,003 → 698 | b9b46cda51df → 78e4cfe68697 | 35.5 | 79.9% | instruction order, priority fee, average tip |
Cw52nP…agc7Wp | 871 → 1,663 | 5def2e6759e4 → 6341f953d383 | 30.7 | 93.1% | instruction order, preferred tip route, priority fee |
Stable controls
Low-change wallets provide a control group: the same feature set can remain recognizably stable across windows despite different trades and tokens.
| Wallet | Transactions, old → recent | Fingerprint, old → recent | Change score | Within-window stability |
|---|---|---|---|---|
3Vggwn…diSVF1 | 1,726 → 1,818 | 65b8f9ddc4fb → 65b8f9ddc4fb | 0.0 | 83.3% |
AowrJa…fNkQJU | 642 → 751 | d556941a1405 → d556941a1405 | 0.1 | 82.7% |
5XEj4q…TuaUVG | 1,016 → 715 | 43201e2490da → 0ccf30c83982 | 0.1 | 91.3% |
4adom2…6KJ9sL | 2,739 → 1,178 | 0f856981353a → 0f856981353a | 0.2 | 94.1% |
9999hu…y9qqRj | 2,057 → 1,539 | d42491748a3c → d42491748a3c | 0.2 | 82.7% |
How to use the signal
Treat a change as a circuit breaker for copytrading: pause, compare the new transaction fingerprint with the wallet's prior baseline, and require fresh performance evidence before following it again. Combining this signal with funding flows, token outcomes, and position sizing is stronger than any single fingerprint component.
View the bounded SQL sample and the fingerprint calculation.