PUMP RADAR · PAPER BOT · ARCHIVE
Paper bot
Our own trading bot, run on paper. It followed every call of our Telegram bot and traded it the way a live bot would: it read the coin's pump.fun bonding curve (or its PumpSwap pool after the migration) on-chain every 5 seconds while it held a coin or was about to trade it (every 15 seconds otherwise), decided on one read and was filled at the next read's real quote, with the trade's own price impact, pump.fun's 1.25% fee each way and 0.001 SOL per transaction. Nothing was bought. Five strategies, fixed in advance, each started with 3 virtual SOL and traded 0.1 SOL; six more aggressive wallets started with 20 virtual SOL each and traded 1 or 2 SOL.
Our study of 761 past calls found that none of the five base rules made money after costs; the aggressive wallets were not part of it. This page measured them again, at real prices and live, losses included, and the answer was the same: every one of the eleven wallets lost. The book was closed on 26 Oct 2026 and the numbers below are final. It was an experiment, not a promise of profit and not trading advice.
Loading the bot…
Aggressive wallets · 20 SOL each
Open positions (UTC)
Day by day (UTC, judged trades)
Latest trades (UTC)
Each of the five base strategies put 0.1 SOL in a trade from a 3 SOL wallet, at most 3 at once and one per coin; after losing 0.5 SOL in closed trades in a UTC day it stopped entering until the next day. The aggressive wallets started with 20 SOL and traded 1 or 2 SOL, up to 10 positions at once; four of them stopped for the day after losing 5 SOL, the runner and the scalp never stopped. Each wallet paid its own price impact, but the wallets did not affect each other: in reality, several wallets buying the same coin at the same moment would push each other's price. Times count from the moment a call was published. A trade is shown but not judged when its entry was late, when the bot may have missed the signal, when the coin could not be read to the end, when the bot's reads stopped (a server restart, a provider outage) at any time from a minute before its entry to two minutes after its maximum hold, or when it was still open at the archive and was closed at its last known price; the rule looks only at those fixed facts, never at how the trade went, so it cannot favour winners or losers. Not-judged trades still count in the SOL balance. A strategy would only have been considered for real money after at least 150 judged trades with an average above zero, a 90% confidence interval above zero, a positive average without its 3 best trades and at least two thirds of full days positive (a full day is a finished UTC day with at least 5 judged trades; at least 3 such days are needed); none got there. Calls on coins paired with a token other than SOL were skipped: they cannot be bought or priced in SOL directly.