Developer Selling and 24-Hour Outcomes: An Observational Cohort Study

Correction and methodology clarification (September 2, 2026): We clarified missing-data, attribution, selection, confounding, auditability and verification disclosures. The cohort and figures are unchanged.

The most asked question in every token's Telegram

"Dev sold — is it over?" is a common question. MemeAssist detects developer selling at analysis time and then tracks the token's next 24 hours automatically. This complete-case observational study compares 626 tokens with detected dev selling in the prior 7 days against 3,153 signal-captured tokens without it in a selected cohort resolved July 22 – August 24, 2026.

Observed 24-hour outcomes in this cohort

In this selected Jul 22–Aug 24, 2026 complete-case observational cohort, the detected-dev-selling group had a 13.6% 24-hour death rate versus 16.0% in the comparison group. Among survivors, its median return was −10.3% versus −2.6%, while its average return was −8.8% versus −3.3%.

These are unadjusted descriptive associations. The study does not establish that developer selling caused the return difference or the death-rate difference, and it does not forecast the outcome of any future token. Developer-wallet attribution can be wrong or incomplete, and the direction and size of resulting attribution bias are unknown.

What the comparison can and cannot answer

• It describes one historical cohort. The figures summarize selected, resolved complete cases from July 22 through August 24, 2026; they are not universal rates.
• It does not estimate a causal effect. The groups were not adjusted for liquidity, market cap, token age, creator history or other risk signals.
• It does not provide a trading instruction or token-level forecast. The aggregate comparison cannot determine whether a particular token will fall, remain quoted, or stop quoting over the next 24 hours.
• The flag has measurement limits. Wallet attribution may miss routed sales or misclassify ambiguous ownership, with unknown direction and size of bias. Our guide to checking whether the developer is dumping explains the wallet-flow inputs.

Returns and death rates are separate outcomes

The table reports two different endpoints: whether a token stopped quoting within 24 hours and, for survivors only, its 24-hour return. In this selected complete-case cohort, the group with detected dev selling had more-negative survivor return summaries but not a higher observed death rate. Neither comparison supports a causal explanation, a universal rule, or a decision about an individual token.

Methodology

Cohort: 3,779 outcomes resolved between July 22 and August 24, 2026 where analysis-time signals were captured. "Dev selling" is the engine's detection, at analysis time, that the creator/developer wallet sold within the prior 7 days — before the 24-hour outcome was known. Death rate = tokens whose market stopped quoting entirely within 24 hours over all resolved outcomes in the group; medians and averages are computed among survivors only and reported separately — a dead token is worse than any survivor average. Both groups far exceed our 30-outcome minimum. Tokens without a captured analysis-time signal or without either a usable 24-hour price or confirmed-dead outcome were excluded rather than treated as unchanged or dead.

Study caveat and limits: creator attribution can miss sales routed through intermediary wallets and can misclassify wallets whose ownership is ambiguous. Such errors may place true dev sellers in the comparison group; the direction and size of that bias are unknown. Tokens submitted to MemeAssist are not a random sample of all Solana launches. The groups were not matched or adjusted for market cap, liquidity, token age, creator history or other risk signals, so this study reports an association rather than proof that dev selling caused the later return. No confidence intervals or significance tests were calculated. These period-specific results from a selected, non-random cohort may not generalize to other periods or to all Solana launches. Survivor-only return statistics also omit confirmed-dead tokens and should be read alongside the separate death rate.

Auditability and verification: public row-level records for this historical cohort are not available. The source facts and outcomes were generated from the internal production tracker and were not manually verified row by row. Accordingly, the aggregate tables cannot be independently row-audited from public data.

How MemeAssist reports the signal

MemeAssist's token analysis reports detected developer selling in the Detailed Risk Breakdown alongside the Overall Health Score, Rug Risk Rating, and a plain-English AI Verdict, using the same detection pipeline that produced this study. The flag and this historical aggregate comparison do not determine a future token's outcome.