Flight Recorder.
Preflight verdicts, scored twice: at +24 h and at +7 d. Each counts as fell 90%, reached 3x or neither. One verdict per coin per UTC day is counted; a repeat is kept but not counted. A verdict that could not be scored is counted as "could not be scored", with its reason, never left out.
- Reached 3x: the market cap at the mark is 3x or more the check-time one.
- Fell 90%: the market cap at the mark is below 10% of the check-time one. For a coin that was on the pump.fun bonding curve at check time, the SOL buyers put into the curve fell below 10% of the check-time amount instead: a curve coin's market cap cannot drop below the curve's starting value (about 28 SOL), so it could never fall below 10% of most check-time figures. A curve coin that graduated before the mark is judged on its market cap.
- Hit and miss, fixed in advance: Avoid is a hit when the coin fell 90% and a miss when it reached 3x. Looks clean is a miss when the coin fell 90%; no outcome counts as a hit for it. Careful has no hit or miss. "Reached 3x" is never a hit for any verdict.
- Two samples: checks people chose to run (anyone can pick which coins those are), and a fixed-rule sample the site runs itself every hour on the 2 most-traded running coins. Each has its own tables.
- Shares: a row shows a share only once it has 30 scored verdicts on different coins from 7 different days, always with its count and a 95% interval. These are counts of outcomes, not a hit rate and not a measure of how often PumpPilot is right.
Loading…
Latest scored marks
Newest first, hits and misses alike, from both samples. Each market cap names its source; "source changed" means the check-time source had no figure at the mark and the other one was used. Each coin links to its page on pump.fun and Solscan, so every figure can be checked.
| Mark | Coin | Verdict (UTC) | Then | At the mark | Outcome | Read | Check |
|---|
Past outcomes are history, not a forecast. Not financial advice. In Claude: track_record. Raw data: /api/record.