own data · closes two days after the cohort day
Curve stall rate
24.1% of pump.fun tokens that first reached 80% of the bonding curve on 21 Sep 2026 (UTC) did not graduate within 24 hours: 367 of 1,526 tokens.
A stalled curve came close to graduation and then stopped. The rate uses the day a token first reached 80%, so each token counts once.
By launchpad · newest complete day
| Launchpad | Stall rate | 7-day pooled |
|---|---|---|
| pump.fun | 24.1% | 26.4% |
| Raydium LaunchLab | 57.4% | 47.8% |
| four.meme | 100.0% | 100.0% |
| Meteora DBC | 0.08% | 0.12% |
Fewer than 5 tokens on the newest day, so no rate is shown: Jupiter Studio, Bags, LetsBONK.fun, Moonshot, Virtuals Protocol. Samples come from platform feeds, so a pad's row covers the curves the engine saw, not every curve on the chain. Pads without curve samples (Pons, Flap.sh) have no row. Download: JSON · CSV. Definition: methodology.
Frequently asked questions
What is the curve stall rate?
Tokens that first reached 80% of the bonding curve on a UTC day and did not graduate within 24 hours, divided by all tokens that first reached 80% that day.
Why does four.meme show such a high rate?
A graduation the engine did not observe counts as a stall. The four.meme feed shows many curves near the top but few graduations, so its rate is an upper bound.
How far back does the series go?
Curve samples and other row-level token data are deleted after 12 to 14 days, so the daily values are stored as they close and are kept permanently. The series starts on 4 Sep 2026, the first full engine day.