2026-10-14

Trench lessons — 2026-10-14

Key takeaways
  • Curve-skip PumpSwap pools: dev-opened pools that look like graduations, 99% get drained
  • Same study, smaller points
  • TrenchBay test launch (Oct 9, updated Oct 13): first-second snipers sell into each other

Curve-skip PumpSwap pools: dev-opened pools that look like graduations, 99% get drainedevidence: A dataset

  • What Bitquery found (launchpad-year study, verified Oct 10; Oct 30 2025 – Jul 5 2026):
  • 451,402 PumpSwap pools were opened by a dev for a brand-new coin that never touched a pump curve. They look like graduations: median 85 SOL, the same as a real graduation pool.
  • The opener holds the whole supply and the withdrawable LP. 99% were emptied by that dev, median 12 minutes after opening, 87% within an hour. The dev left with a median +40 SOL.
  • Real graduation pools: 98% had nothing taken out.
  • Volume of these fell to about a third of the Feb–Mar rate in Jul–Sep, but they're still happening.
  • PumpSwap Pool layout after the 8-byte discriminator: pool_bump u8 @8, index u16 @9, creator @11, base_mint @43, quote_mint @75, lp_mint @107.
  • A migration pool's creator is pump's pool_authority PDA (seeds ["pool-authority", mint] under the pump program) and the pool PDA uses index 0. For a curve-skip pool, creator is the dev wallet.

Same study, smaller pointsevidence: A

  • Graduation is a weaker signal now: a third of Sep pump.fun graduations took ≤3 curve trades (2% in Nov), so a few big buys fill the curve at once. On Meteora DBC, 35% of coins graduate on a single trade.
  • Base rate: half of pump.fun launches are dead within ~3 min, 79% never trade after day one, and 1 in 437 has 10+ trades in week 5. Graduated coins: 43% still trade in week 5, versus 1% for the rest.
  • Meteora DBC coins with transfer hooks (added Jun 2026) don't show up in Bitquery's trade data at all.

TrenchBay test launch (Oct 9, updated Oct 13): first-second snipers sell into each otherevidence: C, n=1

  • On their own test coin, 7 outside wallets bought 3.97 SOL in blocks +1 to +3, and 6 sold within about 3 s. Only the first seller made money (+12%). The other five lost 2–14%, because pump's 1.25% fee on each side means a round trip needs about +2.5% just to break even. The next real buyer came in about 3 s after launch.
  • TrenchBay 1.1.35 ships a launcher-side "sniper dump" that sells everything when an outside buy above a set size lands in the first seconds. Maran's X thread (Oct 14) lists the bundler tools in common use (jtools, solbundler, smithii, trenchbay).

Sources

  1. https://bitquery.io/investigations/solana-launchpad-year
  2. https://github.com/pump-fun/pump-public-docs/blob/main/idl/pump_amm.json
  3. https://github.com/pump-fun/pump-public-docs/blob/main/idl/pump.json
  4. https://trenchbay.com/blog/what-sniper-bots-do-in-the-first-seconds-of-a-launch/
  5. https://x.com/TheMaran/status/2110375832716038633
  6. https://github.com/pump-fun/pump-public-docs/commit/639b87fb45c3adeffa6f7c4a1bf0cb85e392bf22
  7. https://t.me/s/pump_tech_updates

Evidence grades: A = large dataset, B = official docs, C = vendor/thread/small sample. Research notes for education only. Not financial advice.