SmugCalls · questions, answered with numbers
Ask the frog
Every answer below carries its denominator. That is the whole difference between a statistic and a headline — and the reason published graduation rates disagree with each other by a factor of five.
pump.fun figures from 2026-07-06 to 2026-08-05 · call figures from 8,790 published calls
What percentage of pump.fun tokens graduate?
SmugCalls measured 825,123 launches over 30 days: 2.62% graduated. The number only means something next to its denominator.
| counted from | launches | graduated |
|---|---|---|
| every launch | 825,123 | 2.62% |
| at least one trade | 697,882 | 2.98% |
| reached 1 SOL in the curve | 468,923 | 4.15% |
| reached 5 SOL | 362,552 | 5.15% |
| reached 10 SOL | 264,443 | 6.71% |
| reached 25 SOL | 128,330 | 12.40% |
Same month, same chain, same definition of graduation. Only the denominator moves, and it moves the answer by 4.7x — from 2.62% to 12.40%. Window: 2026-07-06 to 2026-08-05.
Why do published pump.fun graduation rates range from 0.2% to 6%?
Because they count different denominators, and most never say which. SmugCalls found the published figures are nearly all defensible — they simply answer different questions.
A launch with zero trades is still a launch. Counting it gives 2.62%. Dropping it gives 2.98%. Requiring the token to reach 25 SOL in the curve first gives 12.40%. Nobody is lying; the denominator is just missing from the headline.
Day-to-day variance adds to the confusion: across 31 days in the same window the daily rate ran from 1.43% to 3.62%. Quote a single day and you can support almost any claim. Full method and the daily series.
How long does it take a pump.fun token to graduate?
Median 5.1 minutes for the ones that make it, measured by SmugCalls across 21,356 graduations. 81% are done inside the first hour.
The tail is long: the 90th percentile sits at 557 minutes, about 9.3 hours. So a token still on the curve a day later is not "about to graduate" — it is in the slowest tenth. Graduation is mostly a first-hour event.
Does the creator's own buy predict graduation?
Yes, but not the way people assume — the relation is not monotonic. SmugCalls measured 0.34% graduation when the creator bought nothing, against 5.70% at 0.05–0.2 SOL.
| creator buy | launches | graduated |
|---|---|---|
| 0 SOL | 125,081 | 0.34% |
| 0–0.05 SOL | 123,729 | 2.05% |
| 0.05–0.2 SOL | 145,476 | 5.70% |
| 0.2–0.5 SOL | 82,763 | 2.32% |
| 0.5–1 SOL | 77,161 | 1.45% |
| 1–2 SOL | 76,142 | 1.72% |
| 2–5 SOL | 156,475 | 1.89% |
| >5 SOL | 38,296 | 7.98% |
Two separate populations sit in this table. A creator who buys nothing at all graduates 0.34% of the time — the worst band by a wide margin, and a fair proxy for "nobody is home". A small buy of 0.05–0.2 SOL does best among modest sizes at 5.70%. The largest band, above 5 SOL, reaches 7.98%, but that is a different animal: capital that size is buying the outcome, not predicting it. Reading this as one smooth curve is the mistake.
What share of SmugCalls calls reach 2x?
35.1% of 8,790 published calls reached 2x or better; the median call peaked at 1.57x. Misses are in the file.
| peak reached | share of calls |
|---|---|
| 1.5x or better | 53.1% |
| 2x or better | 35.1% |
| 3x or better | 21.6% |
| 5x or better | 11.5% |
| 10x or better | 5.0% |
| 50x or better | 0.7% |
| 100x or better | 0.3% |
Peak means the highest market cap after the call, not an exit price — nobody sells the top. The median is the honest headline: half of all calls never doubled. Every call, with the misses.
How can you verify a call channel is not deleting its losing calls?
Ask for a hash that predates the outcome. SmugCalls publishes the full file, its SHA-256, and a Bitcoin timestamp — a deleted call breaks the hash.
Screenshots prove nothing, and a channel can quietly remove a bad call before you look. The check is cheap: download calls.jsonl, hash it yourself, and compare against attest.json. The receipts are stamped into the Bitcoin blockchain through OpenTimestamps, so the record cannot be back-dated and cannot be edited after the fact.
That is the whole claim. Not "trust us" — check us.
Every call on record · raw dataset · how the graduation numbers were measured · home
Graduation figures come from a 30-day census of the bonding curve, every denominator spelled out. Call figures are recomputed from the public file on every build, so this page cannot drift from the record it cites. Peak means the highest market cap reached after a call, not an exit price.