---
name: rh-early-cohort-overlap
description: The sub-$1M early-buyer overlap screen across nine Robinhood Chain tokens (PONS, DELTA, HOOKR, BOW, LOCK, CASHCAT, VYNEX, TWO, GG), its method, its results, and the airdrop trap it exposed. Read before any cross-token wallet-overlap screen on RH chain.
sources: [cowork]
aliases: [sub-1m-overlap, nine-token-overlap, early-cohort-screen]
---

# Sub-$1M overlap screen, 29 Aug 2026

- [stated] Ben asked: find wallets that were early (under $1M market cap) in $PONS $DELTA $HOOKR $BOW $LOCK $CASHCAT $VYNEX $TWO $GG, rank by number of buys, group by coverage
- Artifact: https://claude.ai/code/artifact/cd5d9d87-5e6b-43c6-8e01-171fc9d3d23d · CSV early_wallets.csv (213 wallets at 3/9+)

## Result
Nobody is 9/9 or even 7/9. Max is 6/9 (2 wallets), then 5/9 (9), 4/9 (48), 3/9 (154).
- THE TRAP: ranked raw, one wallet showed 9/9 with 979 "buys" and two showed 8/9. All three bought nothing. They were airdropped the basket by a multisend network. Filter that fixes it: on a real buy the token recipient IS the tx signer. That one test deleted the entire 7, 8 and 9 tier (4,109 of 30,896 receives dropped).
- A farming network sprays these exact nine names at the same wallet list. Source contract 0x39b38686a19836ac10162c490e4558e120cbbe5f alone pushed 1,944 times across seven of them; 0xb92fe925dc43a0ecde6c8b1a2709c170ec4fff4f spans eight; 40+ distinct addresses sign the multisends. ANY inbound-transfer overlap screen on RH chain returns this network first, for any basket.
- CASHCAT is the binding constraint: 491 sub-$1M buyers, NOT ONE reaches 4/9, best pair is BOW at 8 shared wallets. PONS: 802 early buyers, only 16 reach 4/9. The repeat cohort lives entirely in the August names.
- Core cluster by shared early buyers: DELTA+HOOKR 163, HOOKR+BOW 152, DELTA+BOW 142, DELTA+LOCK 115.
- 6/9: 0x4123ffc81579e38dfff74b75d8bdf21c41e62237 (the already-flagged farmer; $2,141 of dust now $121,448, never sells) and 0xf703fd64093b50797abdc9e450632240fa2ba5d4 ($22,683, ~100 tx/day, out of four of six).
- Best single find, 4/9: 0x711281c1b26aaed86e40e4caaf76c1962b45e161. $4,753 deployed, 25 tx/day (human range), DELTA $59k / HOOKR $206k / BOW $224k / TWO $352k entries, still holds 74% of TWO and 32% of DELTA, remaining early bag $46,790.
- Almost everyone else at 4/9 and 5/9 is flat out of every name. The two biggest 5/9 spends (0x6b6e4c1afd72024b1749a1322a524557f033ac04 $30,465 and 0x4526aa01b19ff960fa747b7b963747c3432157c7 $25,652) run 180-280 tx/day and hold nothing.
- Cross-reference: none of the previously audited wallets (0x7e3Ba68C, 0xFF2bCd8E, 0xf974E138, 0x194d98d1, 0x56E95b47, 0x4d9aA6d0, 0x235Bb778) reach above 1/9 here. Being a sub-$1M repeat buyer and being a profitable trader are different populations.

## Follow-up 29 Aug: "which of these wallets makes the most sense to follow" — full priced ledgers on 7 finalists
- WINNER, the only one that clears Ben's consistency bar: **0x711281c1b26aaed86e40e4caaf76c1962b45e161**. Deployed $119,558, realised $77,613, open $91,753, NET +$169,366 (+142%), win 22/40, THREE winners over $10k, top name 40% of profit, traded within the last hour, plain EOA, nonce ~2,975.
  Book: CASHCAT +$68,321 (entry $14.4M mcap, 64% sold, $40,165 still open), DELTA +$55,439 (entry $704k, 70% sold, $38,827 open), PONS +$15,280 (100% sold), HOOKR +$9,622 (entry $820k, 100% sold), FRONG +$6,197. Worst loser ever is PACK at -$2,195; nine losers, none above $2,200. Banks most of a winner, leaves a tail running.
  Its live small end (the undecided positions): PDEX 0x5c9a6a0353f1edc67b22f2cdcfe9294f5a889b60 entry $184k nothing sold, GG entry $2.09M nothing sold, TWO entry $266k 26% sold, DTF 32% sold. CASHCAT and DELTA are already 14x and 24x, so copying in today buys the tail.
  Its names overlap DineroDom's disclosed book closely (CASHCAT, PONS, DELTA, HOOKR, FRONG->PONS, DTF, HOOD10, STACK, BOW). Worth a look in that open hunt, NOT confirmed.
- THE ZERO-COST-BASIS TRAP, the thing that decided this. 0xad16a1ebe5f5aa5dfc3f7c40635b6dd19899f435 looked like the winner at $201,531 net until you split bought from received: $151,764 came from tokens that arrived by wallet-to-wallet transfer with no cost basis (WISHBONE, PIPEDOG, BRODIE, VLAD, UP, WOOF, MANCER, all 100% gifted). It is a sell-side wallet emptying someone else's inventory. Clean net $39,329 on $432,796, one winner, 76% in one name. ALWAYS exclude tokens whose inbound is >25% EOA-transfer before ranking a wallet.
- Also failed: 0xbd66c83f ($529k deployed for $718 net, realised negative), 0x578afe9e ($89,747 deployed, zero net clean), 0x9b9f19ad (+$14,087, no $10k winner), 0x964bf6bf (+$12,464, 58% one name), 0x1440131e (+$3,970 across 450 tokens).
- FOCUS METRIC that pre-sorted the field cheaply: count of tokens the wallet both received AND sent. Machine wallets run 900 to 6,250; the survivor and the near-misses run 129 to 490. Ben's earlier bot rejection (0x311aaC4d) was at 815.
- METHOD UPGRADE: eth_getLogs with topics=[Transfer, walletPadded] and NO address filter returns a wallet's ENTIRE chain history in about 1 second, full range. Two calls per wallet (one per topic position) gives a complete ledger. Recursive range-halving on timeout is required for busy wallets. This makes full ledgers cheap; do not go back to per-token scanning.
- Valuing legs from the QUOTE side does not work on this chain: most swaps route native ETH in Uniswap v4, so there is no ERC-20 quote leg (1,568 of 1,593 txs for one wallet were unpriceable that way). Fall back to the token's GT hourly close at the block timestamp.

## Cluster A and 0x08a16898 — full ledgers, 30 Aug 2026 (both REJECTED as follows)
- [stated] Ben asked, off the ACRE holder sweep in [[acre-lending]]: what else has Cluster A bought, who are they, is it worth following, and the same for 0x08a16898f9b85308d1316984c1804abc663d86c6
- CLUSTER A = 0xf17861250703231dacb1119c4a87324ef950c97d, 0x429fa4e0282278348d26ecbc235f0ef475e0e590, 0xf9c13756b50132623f2529da8f4478359c0f6e0e. ONE OPERATOR, proven six ways: funded by the Robinhood hot wallet 0xf70da97812cb96acdf810712aa562db8dfa3dbef within 60 seconds of each other on 21 Aug 22:45-22:46 with 24.82 / 25.15 / 25.79 ETH; identical 63-token bought set (set equality, not overlap); 58 of the 63 first-bought in the same block; two of them spent EXACTLY 20.5667 ETH to routers; identical 232 / 233 / 234 tx counts and 28 tx/day
- Do NOT cite the 145 shared transaction hashes as proof of one operator. I checked: zero of the 101 buy legs are shared. All 145 are spam multisends from the airdrop network (0xdff7c89bb3019eb383d201422039cc107791921d 108 legs, 0x22ae7e8c970eea0391b016253ed53470df1a0d53 31, 0x315b65cb3d038fa549fa72e69eaed6adebaa0a09 26). Being on the same spray list is a weak signal; the identical buy set is the strong one
- Cluster A behaviour: NOT snipers. Median 52 hours from a token's first hourly bar to their first buy (p25 17h, p75 102h). 54 of 63 names bought on 29-30 Aug alone, so the book is one day old. They exit only 29% of names and when they do, median hold 1.1h. 101 buy legs against 49 sell legs
- Cluster A result, balance-sheet method (deposits vs ETH held plus bought tokens, no price series needed): -14%, -13%, -9%. Each was funded about $61-63k and now carries $52-58k excluding the free airdrop bags. They have withdrawn nothing. Biggest holdings VYNEX, QUOTIENT, FLR, PINDEX, COAT, ACRE. VERDICT: a fresh spray-and-hold bot with a one-day-old untested book, currently down. Do not follow
- 0x08a16898f9b85308d1316984c1804abc663d86c6: live since 8 Jul, 52 days, 1,750+ txs, 34 tx/day, 217 names bought, 328 buy legs against 436 sell legs. Median 2.3h from a token's first bar to its buy (p25 1.0h) and median 6 MINUTES held before its last sell. A fast scalper, not a position trader. Genuinely multichain: 5.53 ETH and nonce 53 on Ethereum, 3.22 ETH and nonce 182 on Base, and it bridges ETH in and out of RH chain constantly through Relay escrow contracts (0x5f281e7d3094a5f44f107c30514031cc959fccc4, 0xda95e31bba4b5c33137b9123eed0e760c4843482, 0x61eb118d689ffe39ad9c449148a86e922d7365f8)
- 0x08a16898 result: balance sheet says +$99k on $147k deposited (+68%) — it bridged 78.35 ETH OUT against 19.26 ETH deposited directly plus 40.48 ETH bridged in. Its four largest single bets ALL LOST: AI 11.00 ETH in / 7.21 out, MGEN 5.05 / 1.37, CLIPPY 3.46 / 0.68, DJT 3.20 / 1.64. The money is a long tail of six-minute flips. Win rate 25% on the priceable subset. VERDICT: probably profitable but it is a high-frequency scalper with a 25% hit rate; there is nothing here a human can copy
- CORRECTION to what I told Ben on the ACRE sweep: I called 0x08a16898 "the strongest single profile" on the strength of its ACRE entry being up 2.85x. That was one unrealised position, not the book. Its book is 217 names and its four biggest bets are all losers
- METHOD, and the trap that ate two hours. On RH chain there are THREE ways to value a swap leg and only one is safe per situation. (1) tx.value from the explorer txlist is exact for buys and had 100% coverage. (2) Internal transactions from routers give sell proceeds but blockscout's internal index is INCOMPLETE: 86% coverage on cluster A, only 50% on the busy wallet, and it also returns reverted legs carrying isError=1 which must be dropped or they inflate proceeds. (3) Decoding the Uniswap v4 Swap event at the PoolManager (topic 0x40e9cecb..., six words, amount0/amount1 as signed 256) LOOKS exact but breaks silently on token/token pools where amount0 is not ETH: it produced a $937k cost basis for a wallet funded with 25 ETH. Do not use method 3 without proving currency0 is address(0)
- THE ONE THAT ALWAYS WORKS on this chain: the balance sheet. Deposits in (external txs plus internal from non-router sources) versus withdrawals out plus native ETH held plus current spot value of BOUGHT tokens, with received-only bags excluded. No price history, no swap decoding, and the router legs cancel. Use it first and use the priced ledger only to attribute the P&L to names
- The GT-price ledger is optimistic on this chain: it marks open positions at spot and drops mispriced pools, which skews toward winners. It scored cluster A at +15 to +21% where the balance sheet says -9 to -14%. Trust the balance sheet on the total, the priced ledger only on the ranking

## Method notes worth keeping
- All nine have fixed 1,000,000,000 supply, so the $1M line is exactly $0.001 per token.
- $1M crossings: CASHCAT 30 Jun 21:55 · PONS 14 Jul 02:02 · DELTA 7 Aug 01:04 · HOOKR 7 Aug 14:58 · BOW 11 Aug 23:19 · VYNEX 27 Aug 14:51 · GG 29 Aug 01:31 · TWO 29 Aug 10:23 · LOCK never (still ~$400k).
- MISPRICED-POOL TRAP: GeckoTerminal reports garbage USD for pools quoted against tokenised equities (BOW/SPY showed $775 BILLION mcap, GG/GLD $416B). First pass took the min crossing across pools and picked those, putting BOW's cutoff three days early and GG's ninety minutes early. FIX: only trust a pool whose latest close reconciles with DexScreener's current market cap within 4x.
- GT hourly OHLCV limit=1000 only reaches ~41 days back. For older tokens use DAILY to bracket, then hourly with before_timestamp, then minute.
- rpc.mainnet.chain.robinhood.com: 403s from urllib without a browser User-Agent, needs one; eth_getLogs times out past ~200k blocks (100k chunks work), 429s above ~6 concurrent; NO archive state (historical eth_getCode fails), but block headers are available so binary search on timestamp works.
- Block time is NOT constant on this chain: ~10k blocks/day in June, ~870k/day by late August. Never estimate a block from a linear rate, always binary search.