Working Paper · February 2026Findings
The Study

The Study
An Empirical Analysis of Sell Timing on pump.fun

Do coins you sell really pump harder than coins you hold?

co_numina · Independent Research
Data: Helius RPC, pump.fun API, DexScreener · Chain: Solana
Abstract

We investigate the widely-reported perception among memecoin traders that tokens consistently appreciate after being sold — a phenomenon termed “the curse.” Using on-chain swap data from Solana wallets trading on pump.fun, we compare each sell price against the token's all-time high (ATH), controlling for temporal ordering by only measuring ATH events that occurred after the sell transaction.

Our methodology filters for pump.fun-native tokens, excludes sub-0.1 SOL micro-sells to remove bonding curve noise, and sources ATH data directly from pump.fun'sath_market_cap andath_market_cap_timestamp fields. This approach isolates genuine missed opportunities from exits that occurred post-peak.

Origin
The hypothesis originated from a widely-shared observation on Crypto Twitter — the persistent feeling that every token pumps immediately after you sell it. These posts captured the sentiment that motivated this study:
Ban
@Bancrypto__

Roundtrip monthly salaries on 4 coins cus I don't sell the dip which 3 seconds later turns out to be beginning of a -90% candle.

Sells the first dip on the 5th coins, proceeds to be the daily millie runner.

Who is gonna do the study

Ban@Bancrypto__ · Feb 16, 2026
Someone seriously need to do a study on coins you hold and believe in going down 80% while coins you sell go up 100x, this shit can't be skill issue atp, it's gotta be something else
4:36 PM · Feb 17, 2026
Methodology
For each wallet submitted, the following pipeline is executed:
  1. Retrieve up to 1,000 recent transactions via the Helius enhanced transactions API
  2. Classify each transaction as BUY or SELL based on token flow direction
  3. Filter to pump.fun-native tokens (mint address suffix pump)
  4. Exclude micro-sells below 0.1 SOL to remove bonding curve exploration noise
  5. Query ath_market_cap and timestamp from the pump.fun coins API
  6. For each sell: if ath_timestamp > sell_timestamp, compare sell price to ATH price; otherwise compare to current price
  7. Compute Curse Score = mean(sell → ATH gap) − mean(hold performance)
Limitations
[1]ATH data represents the token's lifetime peak, not necessarily the peak within a bounded post-sell window.
[2]Tokens with no DexScreener or pump.fun listing are excluded from analysis (dead/delisted tokens).
[3]Aggregator-routed swaps (Jupiter, Meteora) may not be captured if Helius classifies them differently.
[4]The 1,000 transaction limit may not capture full history for highly active wallets.
Interactive Analysis
Submit a Solana wallet address to generate an individual study report. All data is sourced on-chain and verifiable via the linked explorers. Analysis typically completes in 15–30 seconds.