In June 2026 our memecoin sniping program showed a gain of 8.88 SOL on its own scoreboard. Over the same 205 trades, the wallet lost 3.04 SOL. The scoreboard was off by 11.92 SOL, and it pointed the wrong way.
How our own records got it wrong
The program wrote down each sale at the price it was quoted just before selling, not at what the wallet received. On a normal sale the two are close. On a token that had just collapsed, the quote still looked like a sale, while the wallet got back almost nothing. For three days the balance fell with no withdrawals, and the records kept showing wins.
On the 130 losing trades alone, the records showed +3.263 SOL. The wallet showed -4.134 SOL.
What the wallet said
We took the 205 trades that had both a buy and a sell on the chain and read the wallet balance before and after each one, tips, fees and slippage included.
| Trades checked | 205 |
| Trades that ended above zero | 75 (36.6%) |
| SOL spent on buys | 10.43 |
| SOL back from sells | 7.39 |
| Result in the wallet | -3.04 SOL |
| Result in our records | +8.88 SOL |
| Signature fees | 0.0094 SOL |
Where the loss came from
55 emergency exits from tokens that collapsed cost 2.977 SOL, 98% of the loss. The other 150 trades together lost 0.060 SOL. Signature fees were 0.0094 SOL, so fees were not the leak. Collapsing tokens were.
The fix
In June 2026 we changed our scoreboard to read each trade's result from the wallet's own balance change in the confirmation transaction. We stopped tuning anything from the quoted price.
We fixed how we measure. The strategy still lost, and we stopped running it on 2026-08-23.
How to check your own numbers
- Open the buy and the sell transaction of one trade on a block explorer.
- Note your wallet's SOL balance before and after each one.
- Compare the difference with what your tool logged. If they disagree, the chain is right.
What we run now
Today we run paper tests, on a sample of pump.fun launches delivered by PumpPortal. They are modelled from reserves read on the chain: each exit is priced as a sale against the token's real reserves at that moment, with fees, and every loser stays in the count. They are not wallet results, and our results page says so.
From our own records, June to August 2026: program journal, section 76 (2026-06-19), and the park entry of 2026-08-23. Figures are wallet balance changes read from the chain.
Data from public sources and our own records. Not financial advice.