Two in a thousand graduate, eight times more with a channel
A survival analysis of 832,941 token launches puts a number on something the market treats as folklore.
2 minAlgorithmic & AI Trading
Arati Uday Kamat's paper applies survival analysis to 832,941 token launches on pump.fun over May and June 2026. In the fast-regime window studied, the graduation rate is 0.198%, with a Wilson 95% confidence interval of 0.189% to 0.208% and a steady-state figure of 0.207%.
The social-presence result is the one that will get quoted. Tokens with a Telegram channel graduated at 1.485% against 0.166% without — a lift of 8.94 times. The strongest single predictor, though, was initial market capitalisation above 30 SOL. The Cox model reports a concordance of 0.858, stable under bootstrap and under a time-window split.
Read the limitation before the headline
The measurement window is roughly six minutes after launch. The author is explicit that this makes the reported graduation rate a lower bound on what would be observed over 24 hours, not a complete estimate. Anyone lifting the 0.198% figure into a model needs to carry that qualification with it.
The 8.94-times lift is also association, not mechanism. A team that sets up a Telegram channel before launch has already done more work than one that does not, and the paper measures the correlation rather than isolating what the channel itself contributes.
What is genuinely useful here is the denominator. Discussions of launchpad outcomes usually cite survivors; a study that starts from 832,941 launches and reports how few clear the bar puts the base rate where it can be argued with.
Retold from arXiv. This is a summary in our own words; follow the link for the original reporting.