# Two-Minute-Delay Solana Copy-Trading Test Loses 41% per Token Across 93 Tokens

By Enna Lee

Canonical URL: https://www.tokenpost.com/news/investing/35409
Published: 2026-10-20T07:23:02.000Z
Updated: 2026-10-20T07:23:02.000Z
Section: Investing

> The test entered two minutes after large wallet clusters bought. A smaller set of tightly connected groups gained about 10% per token across 16 tokens.

A test of delayed Solana memecoin copy trading lost 41% per token across 93 tokens, showing that early gains among wallet clusters did not carry over to traders entering after those clusters bought.

The test entered positions two minutes after large clusters bought. Its median result was a 95% decline. Results varied, however: a few tightly connected groups generated profits of about 10% per token across 16 tokens.

The broader review tracked new-token trading on Pump.fun and Stonk from late August. It covered more than 1 million liquidity pools, 7.7 million wallets and 384 million swaps. The large sample allowed the tests to compare early cluster entries with strategies that followed those entries later.

Over two weeks, 22.8 million first buys were measured against a specific sequence: doubling before a 40% drop. That happened in 17% of cases. A simulated strategy that took profit when a position doubled and exited after a 40% decline lost about 19% per trade after costs. The figure does not describe what happened to every buy that failed to double first.

The wallet clusters contained 450 to 2,000 wallets and often bought within one or two minutes of a token’s launch. Between 28% and 46% of tokens linked to those clusters doubled. Cluster entries doubled before a 40% drop in 50% to 60% of cases, compared with 17% for other wallets. Those early entry results did not translate into gains in the two-minute delay test.

Following clusters with positive records still produced an 11% loss per token. A separate backtest following a single wallet lost 20% per trade. The tests measured different ways of choosing trades, but both returned losses for followers.

A neural network increased the share of selected tokens that doubled from about 10% to 26%. Its 771 picks over eight days nevertheless produced a loss on every day tested. Picks lost about 11% apiece, their median result was a 45% decline, and 55% hit a stop loss within 24 hours.

The testing effort cost thousands of dollars for APIs, virtual private servers and modeling. It did not identify a monetizable trading edge despite the stronger early results associated with some wallet clusters.