Bitquery counts 6.85 million sandwich attacks on Solana
Bitquery counts 6,850,144 sandwich attacks on Solana from 30 October 2025 to 7 October 2026. Gross take is 309,010 SOL, $29.7 million before costs. 96% hit Pump.fun.
6.85 million attacks in 11 months
Bitquery counts 6,850,144 sandwich attacks on Solana from 30 October 2025 to 7 October 2026. The trades came from 2,545 bot wallets and caught 2,154,994 wallets.
Gross take was 309,010 SOL. Bitquery prices that at $29.7 million, using the SOL price in the minute of each attack, before fees and tips. 96% of the attacks landed on a Pump.fun launch curve or on PumpSwap.
Note: This is reporting, not investment advice. Pump.fun, PumpSwap, and Jupiter appear because the trades ran through their programs. Bitquery does not treat that as a finding about those companies.
| Figure | Value |
|---|---|
| Window | 30 October 2025 to 7 October 2026 |
| Sandwich attacks | 6,850,144 |
| Bot wallets | 2,545 |
| Wallets caught | 2,154,994 |
| Gross take | 309,010 SOL ($29.7 million) |
| Launch curve or PumpSwap | 96% |
| Per day, November and December 2025 | about 3,400 |
| Per day, September 2026 | 51,900 |
Dollar figures use the SOL price of the minute. Venue fees, network fees, and tips are not deducted.
What counts as a sandwich
A bot wallet buys a token in a pool and soon sells 99% to 101% of that size back into the same pool. In between, another wallet buys the same token at a worse price, for at least 1% of the bot’s size. The round trip fits inside three validators’ turns in a row, about five seconds. The wallet in the middle is not making the same kind of round trip.
A wallet counts as a bot only after at least 100 of these round trips, at least 60% of them profitable, and only if those trades are at least a quarter of its swaps. Bitquery drops round trips that start in a coin’s first 10 slots and treats them as sniping. That set is 3.2 million round trips, worth $44.3 million to the bots, and it sits outside the sandwich total.
The base rules come from the paper “No Place to Hide,” 23 September 2026, by researchers at ETH Zurich, the University of Lisbon, Flashbots, and Category Labs. The 10-slot cutoff is Bitquery’s own. The count uses Bitquery’s records of 10.7 billion successful swaps on Pump.fun, PumpSwap, Raydium, Meteora, Orca, and other pools. Bitquery checked 78 cases on a public Solana node. Order, amounts, and pool matched.
In the paper, the daily average falls to 20,663 after validator delistings in October 2025. Using those rules, Bitquery gets 20,700 a day for November 2025 through June 2026. Over 1 July 2023 to 30 June 2026 the paper counts 28.0 million sandwich attacks on Solana. The window and the filters there are different.
Pump.fun, slippage, and other bots
The daily count was about 3,400 in November and December 2025 and 51,900 in September 2026, about 15 times as many. The median sandwich made the bot $2.33. One in ten made more than $12.
Among Jupiter swaps into Pump.fun coins, trades that allowed 20% to 50% slippage were caught 14 times as often as trades under 3%. Across all coins the gap is wider, because tight settings mostly go to other pools. 39% of the gross take came from other bots’ trades, including snipers. Bitquery puts the other $18.1 million with the remaining wallets.
In a sample of 63 attacks the bots kept 68% of the gross after fees and tips. 18 of the 63 lost money once costs were counted. That sample is not the full set. Where a sandwich had one victim and nothing else in between, Bitquery put the victim’s loss at 1.3 times the bot’s take on PumpSwap and 1.2 times on the launch curve.
Only 6% of the sandwiches buy and sell in the same slot. The September paper says the jitodontfront flag only stops another transaction from going ahead inside the same Jito bundle. After the October 2025 validator delistings, the attacks shifted to sequences that span several slots. Bundles and slippage are covered in MEV on Solana. The figures in this piece are Bitquery’s.
The ten wallets with the largest gross take account for 13% of it. The top 1,000 of the 2,545 wallets account for 95%. Half of the bot wallets were active on seven days or fewer.
Limits of the count
Bitquery calls the total a floor. HumidiFi and Tessera V are missing from the records, and so are about 1.6% of blocks. Wallets that only sometimes make these round trips are left out. An estimate from sample days puts both the attack count and the gross take at about double, and the gross take at $58.6 million to $63.5 million.
In the same window Bitquery counted address poisoning and failed swaps separately.
What to watch
- The $29.7 million figure is gross. A net figure after fees and tips exists only for a sample of 63 attacks
- HumidiFi, Tessera V, and about 1.6% of blocks are missing from this count
- The study ends on 7 October 2026
Sources
- Bitquery — Sandwich attacks on Solana, on-chain figures verified 11 October 2026
- Heimbach, Solmaz, Öz, Ferreira Torres — No Place to Hide, 23 September 2026
- SOLANA·HUB — MEV on Solana · Address poisoning · Failed swaps