Calculation

How Historical Token Behavior and Exit Patterns Work

Token Behavior and Exits are frontend calculations over retained mint history. They classify observable event signatures and do not claim to prove trader or bot intent.

Pages and availability

Both pages require the token's CreateEvent to remain inside the event-cache window. PumpScan disables their navigation when creation is outside that window because creation-relative timing and retained positions cannot be reconstructed reliably.

RoutePurpose
/token/[address]/behaviorHistorical Trading Behavior shows matched token strategies at the selected confidence.
/token/[address]/exitsHistorical Exit Patterns classifies retained sellers as Creator or Trader and shows one primary pattern plus other matches.

History and ordering

  • The frontend reads mint-filtered /v1/events history without a local-day filter, up to 25 pages, and warns when the result is truncated.
  • Failed transactions are excluded before classification.
  • Rows are ordered by slot, transaction index when both rows have one, event time, signature and event index.
  • A browser worker performs the calculation; loading remains active through both fetching and analysis.

Filters and evidence

  • Observed, Estimated and Incomplete are confidence levels, with Observed selected by default. The Behavior page displays matched strategies only.
  • For Mayhem tokens, Hide is enabled by default and recalculates results without the Mayhem wallet's trades.
  • Expanded rows begin with a Mint origin row at 0s ago and show creation-relative Account, Event, Amount, Account Change, and reconstructed real quote reserves before and after each event.
  • Creator self-buy requires at least two explicit creator buys after creation in the first minute. CreateEvent is not counted as a buy.
  • Whale entry uses the retained p95 quote size or at least 5% reserve-derived price impact threshold.

Fee-aware reconstruction

Observed Pump fees sum TradeEvent fee, creator fee and buyback fee. Position cost adds those fees to buy quote and subtracts cashback; sell proceeds subtract those fees from sell quote and add cashback.

For evidence tables, real quote reserves before a buy are reconstructed as reported post-event reserves minus quote amount. Before a sell they are reconstructed as reported post-event reserves plus quote amount. These values explain reserve movement; they are not independently fetched balance snapshots.

Limits

  • Token transfers and historical holder balances are not retained by this event source, so transfer-dependent ownership and aggregator strategies remain unavailable or incomplete.
  • Estimated price, return, flow, curve-progress and reserve-impact classifications depend on retained event fields and can be incomplete at the cache boundary.
  • The backend does not expose a universal bonding-progress percentage. The frontend may estimate curve context retrospectively from the available reserve fields.