---
name: crypto-token-research
description: Systematic crypto token research process, automation, active tokens, and copy-trade targets
sources: [backfill, cowork]
aliases: []
---
- [stated] MANDATORY on every screen, see [[data-access]] for the credential and the detail: query X via twitterapi.io (key is stored there, works direct from the container, never report X as unavailable, never substitute LunarCrush), and produce the numbers via the existing `three-frame-valuation` skill - fundamental anchor, live-product floor, narrative ceiling. Do not invent frames.
- [stated] Developing and refining a systematic token research process across Robinhood Chain and Solana
- [stated] Process order: kill checks first (rug/authority/liquidity) before reading the pitch, then holder + bundle analysis, DexScreener financials with timestamped dual price frames (rolling 24h AND drawdown from ATH), multi-format website verification (never declare empty from one static HTML fetch), product vs. token value capture split (two separate verdicts), X authenticity + smart money flow, metric-based competitor comparison, verdict + "what would change my mind" trigger
- [stated] Volume: always use GeckoTerminal OHLCV candles, not CoinGecko; DexScreener token-level volume is inflated by cross-pool arb
- [stated] Cowork = diligence only, never connect wallet or execute trades
- [stated] Self-improving source registry: when a mention cites a source (podcast, Space, team convo), capture it to Source Registry, score reliability across runs, auto-promote proven sources into the standard checklist
- [stated] Automated screening via Cowork with a daily Bangkok-time digest and hourly Track A/B/C scanner
- [stated] Active tokens researched: STONKBROKER, MANCER, DCR, HOODRAT, USELESS, HYPE, CARDS, TIBBIR, EARN (Robinhood Chain)
- [stated] Typical screening tier for a new low-cap candidate: $5k–15k, weeks-to-months cycle hold (not the $100–500 lottery-ticket size the lowcap-screen skill assumes)
- [stated] Lesson from HOODRAT: always report both rolling 24h change and drawdown from ATH; use GeckoTerminal OHLCV candles for volume, not CoinGecko
- [stated] Screened DTF (Down to Finance, 0xeE5576Fa1Bcaa380e591D01245f406f3f384eb01) and compared against NET (NetNet) on Robinhood Chain; sizing $2.5k–10k, weeks-to-months horizon
- [stated] Tracking X trader kioto (@0xkioto) as a copy-trade target; rates him a great trader and wants to enter ahead of his public calls rather than after them; asked for his wallets to be identified before any bot is built

- kioto's identified EVM trading wallet: 0xAcb1d388357aE2167CC03f8752cA5ad02775D2Db (same smart account on Base, Ethereum, BSC and Robinhood Chain, routed through fomo); Solana wallet still unidentified
- [stated] Following a second copy-trade target, a swing trader at EVM wallet 0x688F9499f748053b14541F5c596c53A1b002F52D; asked for his other wallets and any influencer/X identity to be found. Linked Solana wallet identified via Relay bridge records: 6VPjjT8WSH3Akt4Ro2tj8gaskCChXgES59Mgzdhfv9jY (dormant since 2 Aug 2026). No public identity, ENS, .sol or KOL-leaderboard listing exists for either address. His main venue is Robinhood Chain; he held ~13.0-13.5M STONKBROKER (~$265k) on 26 Aug 2026, a token Ben also holds.
- [stated] Pays for a subscription to @_0xghost_ on X (44k followers, Robinhood Chain and memecoin caller); a third copy-trade / signal target, asked for his wallets to be identified from his posts with emphasis on Robinhood Chain
- kioto (@0xkioto) full wallet set, proved via exact fomo balance and trade-size matches. LIVE: EVM 0xAcb1d388357aE2167CC03f8752cA5ad02775D2Db (EIP-7702 EOA, same address on Ethereum/Base/BSC/Robinhood Chain); Solana trading 6K3KqNrWrfCnZJPhn143QhfKBvwiWejpsDesF2p7aNjY; Solana vault JWTpAZt4s7jYiFBjmBR2K3qdmrK673DctEZhqrmCU3A (where he parks what he actually keeps). DORMANT: HvFaw9Rd24ykCi5D2G72dKUGzMEiLc3YKGQqLYidPCHY, F4v9sj1XSNoSRvxHjSNeM2Gr2DZEX7jkJusPpWnPT2Y, FYwMbJ1uS2nYdRRm9rkRXPNYbYE35k4ZBBDm1qaucbKt
- kioto behaviour: buys 11 minutes to 12 days before posting, but exits within hours to days. His published multiples describe the token's move from his call, not his PnL (ANSEM 1.76x on a claimed 8x, GOBLIN net negative on a claimed 6x). Real conviction signal is a transfer into the JWTpAZt4 vault, not a tweet. fomo's public API returns two decoy addresses for him that have never transacted on any chain
- 0x688F9499 wallet set (the second copy-trade target): PRIMARY 0x688F9499f748053b14541F5c596c53A1b002F52D (Robinhood Chain, main venue). BASE LEG 0xf5483bc0F2f89A22282c57E028d4D6C2F63958a4 (funded 20.04 ETH one-way from the primary on 16 Aug, same routers, both trade QSAVE). SOLANA 6VPjjT8WSH3Akt4Ro2tj8gaskCChXgES59Mgzdhfv9jY and 3zrdZrQLZdwSo87Jt8ep9bBGLeaqY93WxT2hLeuTt6Cs, both dormant. NOT his: 0xD7ADe0AD95A91aca12abe61e65c28f324a5A78F4 (disperser paying 19 different recipients). He uses ghostswap.eth on Ethereum, so wallets funded that way stay invisible to funding-trace
- 0x688F9499 behaviour, measured 2-27 Aug 2026: "buy red sell green" confirmed - mean buy at 0.28 of the trailing 24h range after a -16.1% move, mean sell at 0.84 of range after +18.3%. But his profit comes from size, not cadence: closed round-trips under 2 ETH returned -31.5%, 2-10 ETH -21.9%, over 10 ETH +0.5%; the +115 ETH net for the month came from liquidating pre-August UP and CASHCAT inventory. Copy rule that falls out: ignore his fills under 10 ETH, mirror only 20 ETH+
- kioto entry quality, measured two ways: Solana wallet Jan-Aug (66 closed round-trips, $71k deployed, +8.9%, 36% win rate, median trade -8.1%, median hold 5.5h) and all chains last 16 days (32 closed, $58k deployed, +2.9%, 38% win rate, median -4.1%, median hold 1.4h). Positive only via payoff ratio (avg win $741 vs avg loss $361) and heavy concentration (top 3 = 60% of profit). Size is the filter that works, same as with 0x688F9499: his entries at $3.5k+ returned +21.4% (Solana, 60% win, 106h holds) and +12.1% (16d, 50% win); his sub-$1.5k clips returned -9.5%. Copy rule: mirror only $3,500+ entries into tokens he has not traded before, ignore everything smaller
- 0x688F9499 rescored over its FULL lifetime on Robinhood Chain (corrects the Aug-only read that called UP/CASHCAT "pre-window inventory" — their entries are inside the lifetime ledger, so the profit is real round trips): 118 closed round-trips, $896,769 deployed, +$270,212 (+30.1%), 43% win rate, median trade -5.8%, median hold 0.1h, avg win $6,445 vs avg loss $873 (7.4:1). Size buckets: <$1.5k +2.2%, $1.5k-3.5k -8.3%, $3.5k-10k -17.3%, $10k+ = 15 trades, $702,491 deployed, +$288,114 (+41.0%), 73% win, 10h median hold. Everything under $10k lost $17,901 across 103 trades. Two positions are 99% of net: UP ($75.6k in, $237.6k out, held 27d) and CASHCAT ($322.2k in, $429.0k out, held 39d); excluding those the $10k+ bucket is still +6.3% on $304,669 winning 9 of 13
- Standing screening conclusion across both wallets scored so far: neither is worth mirroring wholesale (win rates 36-43%, negative median trade in every sample). Both are profitable only in their top size bucket. So the lead metric when screening any new wallet is return on the largest entries, not headline PnL or win rate. Ranking: 0x688F9499 first (alert on $10k+ entries), kioto second (alert on $3.5k+ entries, doubles as confirmation for his articles)
- [stated] Explicitly looking for accounts/wallets that identify projects early AND hold them into the wide-attention phase, not flippers — this is the trait he is screening copy-trade targets for
- [stated] Fourth copy-trade / signal target screened: X caller @dontbuytops ("Ferr", formerly DefiFerr, ~13.8k followers, Robinhood Chain + Base + Solana micro-caps). Forensic candidate wallet 0x59472bec28e874a20d7fbc134337a65ae509c037 (EOA, Robinhood Chain main venue, also trades Base) — matched by intersecting buyers across four of his timestamped calls; not cryptographically proven
- 0x688F9499 execution fingerprint, proves the wallet set is one operator running one tool: sells fire as a scripted decaying ladder where each tranche is EXACTLY 0.9000 of the previous (verified to 4dp on Robinhood STONKBROKER 19 Aug and Base QSAVE 15 Aug, two chains two tokens), with a 0.95 decay variant also in use; buys are fixed round clips fired 12-30s apart (Base 3.0/2.0/1.0/10.0 ETH, Robinhood 10/15/20 ETH, Solana 3.0/10.0 SOL). The Base leg 0xf5483bc0 shares both the 3.000 ETH QSAVE clip sequence three days later AND the oddly specific 0.275 ETH clip — that shared non-round number is the strongest single link in the set. Holding period varies by venue, not by style: Base = scalping (QSAVE round-tripped inside 3h repeatedly), Robinhood = swing (STONKBROKER held 8d+), Solana 3zrdZrQ = one 8-minute laddered liquidation of ~209 SOL (~$21k at $101/SOL) then bridged out — it is a liquidation wallet, not the throwaway an earlier read called it
- Practical alerting consequence: because the sell ladder fires 5-15 tranches within minutes, any per-fill alert will spam. Alert on the FIRST tranche of a ladder and on buys only
- [stated] Fifth copy-trade / signal target screened: X caller joji (@metaversejoji, ~264k followers, Robinhood Chain + Solana + BSC memecoins, runs t.me/jojiinnercircle). Asked specifically whether he has private wallets front-running his public fomo trades
- joji wallet set: he self-disclosed two SOLANA wallets on X 19 Aug 2026 — 68XiYbqQedUbV1EboBhtZDtTTzX4vwUBzKB6rbUGPvHj ("fomo wallet, more size and longer hold times") and 525LueqAyZJueCoiisfWy6nyh4MTvmF4X9jSqi6efXJT ("axiom wallet, scalps and fast trades"), with "do not copytrade me, just spectate". 68Xi… confirmed his by an exact 0.290 SOL balance match against his live fomo book. 525L… has NOT signed a transaction since 29 Jul 2026 and ~1,058 of its 1,100 funded SOL left untraced; 42.4 SOL remain. No on-chain link between the two wallets and no high-confidence sibling found (only MEDIUM candidates via shared funder C43G768JEQsVVaFgY1zCMMAKmGARP67sbB2kBYmV8XKi, consistent with an exchange relationship)
- joji needs no hidden wallet — the real read is fomo.family/profile/metaversejoji, which exposes his full live book and every swap with mcap and size. 68Xi… is only the Solana leg of that one fomo account; the Robinhood Chain and BSC legs are separate addresses whose trades all still surface on the same public profile
- joji front-run receipt: b-money (BSC, pair created 22 Aug 20:29 UTC) — he bought 6 clips at $205.7K / $208.1K / $261K / $299K / $321.9K / $796.4K mcap, then posted the CA to 264k followers at "$760K mcap" on 23 Aug 01:39 UTC. Average entry roughly one third of the mcap he published
- joji size-bucket read (the standing lead metric): the book is one trade. Portfolio $445,028 of which DELTA is $298,069 at +1,476%; top 3 = 87%. Cash $2,164, fully deployed. 593 trades since Oct 2025, 2d20h avg hold. Self-reported drawdown was $874k on 18 Aug ($381k fomo + $493k axiom); the DELTA position roughly erases the fomo half but is entirely UNREALISED, and the axiom $493k sits unrecovered in a wallet he stopped using. Not profitable — one open winner in the exact setup he already described losing ("was up almost $400K in worldcup and convinced myself to not sell"; "I rt'd 400, made like 50")
- joji Aug call record from tweet timestamp to 27 Aug: DELTA +5.0x (called 15 Aug, entry honestly stated at ~$700-770k and matches his real basis), website +7.9x, NET +2.6x, b-money +0.14x, HTZ -92%. Calls working while his disclosed wallets are dead is itself the tell about where he trades
- joji dumped ~$16.9k of LAYOOO (at $661.2K and $715.2K mcap) on 26 Aug — a position Ben also holds
- [stated] Confident 0x96b13d47C17d61dC42F2152FA39Cf3acffc7eBfC is @_0xghost_'s main Robinhood Chain wallet; does not think Ghost dumps on followers, and believes his other coins sit in a second wallet still to be identified. Asked for a dedicated real-time bot on that wallet alerting on buys and sells above $1,000
- 0x688F9499 identity, second attempt with the techniques that cracked kioto: still ANONYMOUS and now more firmly so. It is a plain self-custody EOA (no EIP-7702, so not a fomo/social-app user, so no public profile to match trades against), no ENS forward or reverse, no NFTs, zero mentions of the address anywhere on X or the open web, no explorer label. Footprint: Robinhood Chain 2,434 txs and 97.07 ETH native (~$239k), Ethereum 24 txs and 8.63 ETH, Base 170, BSC 133, nothing on Arbitrum/Optimism/HyperEVM/Ink, no Hyperliquid account
- Ruled out: 0x688F9499 is NOT @_0xghost_. TIBBIR is Ghost's most-posted ticker (17 mentions in 3 months) and appears nowhere in this wallet's 627-token Robinhood Chain history; Ghost's first CASHCAT post (9 Jul, calling the on-chain behaviour "a cabal playbook") came a day AFTER this wallet put ~$97k into CASHCAT in four minutes, so Ghost may literally have been describing this wallet
- Also ruled out as personal links: 0xf70da97812CB96acDF810712Aa562db8dfA3dbEF and 0x4cD00E387622C35bDDB9b4c962C136462338BC31, which appear in both kioto's and 0x688F9499's flows, are Relay bridge solver/depository infrastructure, not a shared counterparty
- 0x688F9499 live position 27 Aug 01:00 UTC: fully out of CASHCAT, UP and DEBT (all zero). Only STONKBROKER left, 9,454,579 tokens (~$184k at $0.01946), down from ~13.0-13.5M on 26 Aug. He has been laddering out continuously 23-27 Aug and was still selling at 00:17 UTC, into a +19.8% 24h and +32.9% 6h move. Relevant because Ben holds STONKBROKER
- Alert fleet change 27 Aug 2026 (DONE, live): rhtracker's tracker.py patched (backup tracker.py.bak-20260827-015949) to split min_usd into min_usd_buy/min_usd_sell, read chain_slug from config, and attach a DexScreener "?maker=" inline button that resolves the deepest pair at alert time. rhtracker now watches 0x688F9499 + 0xf5483bc0 + Ansem RH on Robinhood Chain, buys>=$5k, sells any size. New sibling service rhtracker-base (/opt/rh-tracker-base, unit rhtracker-base.service) runs the same code with chain_slug "base" and the SAME Telegram token+chat so both land in one thread; zero chain+wallet overlap so no duplicate alerts. Both in ~/botctl/registry.json and added to watchdog.py UNITS (now 11 units). Fleet is 12 bots, all healthy. The two Solana wallets are NOT covered — tracker.py is an EVM log poller and Solana needs a separate poller
- [stated] Wants every sell tranche from 0x688F9499, no ladder-collapse suppression, despite the scripted 0.90 decay ladder firing 5-15 messages per exit
- 0x688F9499f748053b14541F5c596c53A1b002F52D FINAL lifetime ledger (supersedes the +30.1% read, which missed 98 native-ETH router swaps through 0x8366a39cc670b4001a1121b8f6a443a643e40951 that emit no WETH log): 2,259 real trades, 140 closed positions, $932,021 deployed, $1,195,385 returned, net +$263,364 = +28.3%. Win rate 40% (56/140), median trade -8.1%, median hold 0.1h, avg win $5,887 vs avg loss $790 = 7.5:1 payoff. Size buckets: <$1.5k -3.4% (69 trades), $1.5k-3.5k -11.2% (41), $3.5k-10k -15.4% (15), $10k-50k +2.6% (12, 67% win), >=$50k +61.7% (3 trades, $456,881 deployed, +$281,712, 100% win, 727h median hold). Everything under $10k lost money across 125 of his 140 positions
- 0x688F9499 best trades: UP +$162,025 (+214.3%, $75,619 in, held 30d, 11 Jul to 10 Aug), CASHCAT +$106,797 (+33.1%, $322,203 in, held 44d, 8 Jul to 21 Aug), GWOOD +$12,889 (+21.8%, held 15d), 530A +$7,418, WOOD +$4,985, PLTS +$4,747. Worst: sushicat -$10,617 (-36.3%), NAVEN -$4,423 (-51.9%), DATABEAR -$3,752, wif -$3,262, MINR -$2,967 (-60.9%). Every loser was held under 24h; every big winner over 15 days
- CORRECTION to the "whale on his way out of STONKBROKER" read: he BOUGHT it, 19,235,330 tokens for exactly 130.000 ETH (~$320k) in 11 clips of 5 and 10 ETH between 18 and 26 Aug, and was still buying on 25 and 26 Aug while selling. Position is mid-cycle, not an exit: $316,602 cost, $201,460 already recovered, 9,454,579 tokens (~$184k) still held, so running about +21.8%. What misled me was seeing only the WETH-legged sells plus the balance drop from 13.5M to 9.45M
- 0x688F9499 style summary: two different traders in one wallet. An intraday flipper (125 positions, median 6-minute holds, every size bucket under $10k negative) and a position trader (3 positions over $50k, all winners, held 15 to 44 days). He buys weakness in fixed round ETH clips and sells strength in a 0.90-decay ladder, cuts losers inside a day, holds winners for weeks. Verdict: worth following, but only the $50k+ entries, or $10k+ as the loose filter. Open book now: STONKBROKER (mid-cycle) plus HOOD $21,677 cost since 31 Jul, and dust in FIST, RETRACE, PUG, CUBE
- Wallet-first screen of Robinhood Chain for the "enter early, hold into wide attention" trait (27 Aug 2026), method: 12 tokens that completed the arc (CASHCAT, PONS, AI, STONKBROKER, HMM, INDEX, JUGGERNAUT, DOGO, IF, DELTA, GOOD, HOOKR), eth_getLogs Transfer sweep of each token's first 12h (81,529 logs), top-400 Blockscout holders per token with contracts/PoolManager/lockers/routers/burn stripped, then full per-position ledgers priced at GeckoTerminal daily closes. New lead metric introduced: CAPTURE RATIO = wallet's multiple / the token's multiple from that wallet's weighted-average entry
- The six that survived, ranked: (1) 0x0428ef4923513f73d53fd6c28875574fc7ed0413 — $53,021 deployed, 19.50x, +$980,626; CASHCAT from 2 Jul at $6.90M avg cap (24.96x, 72% capture, 54% sold), PONS from 15 Jul at $9.23M (9.33x, 97% capture, 0.4% sold); 371 txs, $890k book. (2) 0x7e3ba68c49561aae7c23c1d20fef0f1d7615a3ad — PONS from 15 Jul at $9.23M, 9.24x, 96% capture, 9% sold, +$777,385; also in HOOKR's first 12h; 473 txs, $1.64M book, THE ONLY ONE STILL DEPLOYING (bought $14,158 NET on 26 Aug, holds DTF $26,450). (3) 0x44df085447dbebcf69c6675c3b8a795c7fdeb3f4 = fullturbo.eth — $191,406 deployed, 5.37x, +$835,595, sold only 12.6% of CASHCAT and 3.1% of INDEX, two-name book. (4) 0xde52c75041371b5e08d8a50b82c63734850628c1 — biggest absolute, +$2,499,944 on $2.70M, all bought on ONE DAY 5 Aug, 110 lifetime txs, sold 0-5.5%; size-and-patience, not early discovery. (5) 0xfa8e161b288386c38eced61ebdea0693de5184ff — earliest verified entry anywhere: HMM at $2.35M avg cap on 23 Jul, 7.50x, 78% capture, small $292k book. (6) 0x56e95b47a07210f1ea9fa418c7c5dc164c419ac0 — the control/failure mode: found CASHCAT 4 Jul at $6.88M then sold 85.3%, kept 11.5x of an available 34.6x = 33% capture
- REJECTED from that screen, and the trap to remember: 0x2dbaf98620ac5bbe3441f756feff82702d095b1a is a top-400 holder of 9 of the 12 arc tokens with a $6.46M book across 982 positions — the best surface profile on the chain — but 28,113 txs / 107,643 token transfers, 700+ CASHCAT legs from 19 Aug at a $177M average, sold 126% of what it bought, net -$345,415 on the chain's biggest winner. It is a market maker whose inventory looks like a portfolio. Any screen ranking by "holds many winners" surfaces it first
- Rarity finding: of 2,883 distinct EOAs holding top-400 positions across those 12 tokens, only 1 appears in 9, 2 in 6, 4 in 5, 10 in 4, 35 in 3, 282 in 2 and 2,549 in exactly one. Early-and-hold is close to nonexistent on this chain
- Deliverable: artifact "Robinhood Chain Capture Board" at https://claude.ai/code/artifact/22a91cbf-ece0-4883-be8c-1419b1062d39
- Second, independent Robinhood Chain wallet screen (27 Aug 2026), TRADE-level not holder-level, complements the Capture Board: Robinhood Chain runs a Uniswap V4 singleton PoolManager at 0x8366a39cc670b4001a1121b8f6a443a643e40951, so every swap on the chain emits one event with both legs (w0 = ETH signed, w1 = token signed, topic1 = poolId, topic2 = router). Swept STONKBROKER (poolId 0xd33c8fd38b06e989cdbd4dffdefab71c4bdd415b24964c8d69e38ff35b068f92, 133,425 swaps) and PONS (V3 pool 0x10cc6bd38112cac182db90b6a71d8bb5939526ba, 285,783 swaps), kept the 14,292 swaps >=2 ETH, resolved traders by tx sender. Validated: ledger says 128.700 ETH spent on STONKBROKER by 0x688F9499, vs 130.000 measured independently, gap = the 1% pool fee
- Screen traps found: MEV bots own the raw PnL leaderboard (top four had 771k-1,055k lifetime txs); and roughly twenty wallets prefixed 0x4337 entered PONS on the same days at 5-14% of range with 10-30 buys and ZERO sells, identical shape, one operator or team allocation. Filter that works: under 5,000 lifetime Robinhood txs AND >=3 ETH into a single name, which left 90 wallets
- Best wallets found, archetype A (early + held weeks): 0xbe9ca43007d5b87062d21be7168b92f2c3a528b7 (684 txs, PONS 14.4 ETH in ONE buy at 10.6% of range, held 42.3d, never sold, 8.64x; holds STONKBROKER 1,293,905 and PONS 162,653 now); 0xa2c440c71f874cb92bb4de7a6f7ba920ae0e6468 (only 118 lifetime txs, Robinhood-only, STONKBROKER 27.5 ETH 5 buys 0 sells at 12.9% of range held 37.9d 2.81x, holds CASHCAT 154,401 = $35,667); 0x11bc74e240c13f7067529b21dfb9933ff7fcb136 (PONS 121.5 ETH 2.68x 21d + STONKBROKER 35.8 ETH 2.02x 37d); 0xf24022ba0fea5926d30fba8fece6d7a852732989 (PONS 12.7 ETH, 2 buys 0 sells, 42.5d, 10.48x); 0x40f1b62ed060edbbbda9ced30e7f57e04675b9e1 (fomo/7702 user, STONKBROKER 24.5 ETH at 2.8% of range, 29.8d, 11.02x)
- Best wallets found, archetype B (pullback buyer): 0x1f16dcd7118b0b45e40e42b6b73364f5075d110e (155 txs, PONS 77.3 ETH across 15 buys averaging 60% BELOW the trailing 3d high, 12.36x, combined +941 ETH across both tokens, the best in the screen, currently flat so it is waiting); 0x8fabe392df90b06ad610c33f41bb1abd78a5fc20 (STONKBROKER 11.8 ETH at 26.8% of range held 20.8d for 9.88x); 0x0b854b7bb40c48a77e2e5f1b21ffd6b45b6335a4 (PONS 29.2 ETH, 9 buys 2 sells, 36% avg drawdown at entry, 10.50x); 0xeba9b11f0e126f14c19ada2e6b629af827f2c906 (PONS 104.4 ETH, 5.28x, ~522 ETH still on); 0x901037edc735252913aca7ef563769879cc32f97 (PONS 104.1 ETH 3.20x 27d, now flat with 62.47 ETH idle = ~$154k dry powder, so its next buy is the alert worth having)
- Cross-check: the screen independently surfaced @_0xghost_ 0x96b13d47C17d61dC42F2152FA39Cf3acffc7eBfC at STONKBROKER 35 ETH, 2.73x, 34 days, entry at 13% of range, which validates both the screen and Ghost. 0x688F9499 scored 1.21x on STONKBROKER entering at 32% of range, consistent with him being a size-and-discipline trader rather than an early one
- Deliverable: artifact "Robinhood Position Traders" at https://claude.ai/code/artifact/108f8c6c-3766-42a0-873e-0612eac3e8d4. NOT yet done: Solana pass (no singleton AMM, needs a different engine), and tokens beyond STONKBROKER/PONS in this trade-level method
- Full lifetime win-rate ledgers for the six Capture Board wallets (every position where they actually paid for the token, priced leg-by-leg vs GeckoTerminal daily closes, 27 Aug 2026). None is a four-trade fluke and none has a good hit rate: 0x0428ef49 14 positions/50% win/$327,694 deployed/+$1,090,658/median +12.5%/payoff 33:1; 0x7e3ba68c 22/59%/$814,534/+$924,780/median +40.3%/4.8:1; 0x44df0854 (fullturbo.eth) 5/60%/$206,335/+$830,807/median +100.4%/53:1; 0xde52c750 7/71%/$2,752,717/+$2,517,251/median +19.3%/6.0:1; 0xfa8e161b 7/57%/$78,690/+$239,896/median +0.0%/138:1; 0x56e95b47 10/30%/$307,721/+$660,782/median -19.5%/15:1. Third independent confirmation of the standing rule: profit is payoff ratio and one or two concentrated winners, never hit rate
- 7-day flows to 27 Aug 2026 — this reordered the Capture Board. FIVE of six were NET SELLERS into the 26 Aug strength: 0xde52c750 -$379,630 (INDEX $174,842, TENDIES $79,827, UP $48,348, STONKBROKER $32,619), 0x7e3ba68c -$104,599 (sold CASHCAT $138,877 + PONS $87,478 + HMM $48,197, rotated into NET $138,263), 0x56e95b47 -$42,593, 0x0428ef49 -$23,654, 0x44df0854 -$12,797. ONLY 0xfa8e161b was a net buyer: +$73,099 with ZERO sells (added PONS $49,520, INDEX $17,749, HMM $5,830) — the accumulator of the set and the one to watch now. The rejected market maker 0x2dbaf986 was also buying (+$85,529 in the 3h to 02:27 on 27 Aug)
- CORRECTION, supersedes every earlier line in this file that links @_0xghost_ to 0x96b13d47C17d61dC42F2152FA39Cf3acffc7eBfC (including the Capture Board cross-check note): that wallet is NOT his. It holds 0 of the 9 tokens he told subscribers he bought. Ghost's wallet is 0x4123fFC81579e38dFFf74B75D8Bdf21c41e62237, proved 9/9: SIBYL (17 May), SUPERGEMMA (23 May) and PRXVT (19 Jun) all on Base and all still held, plus SESTRI, FIH, AI, QUOTRON, PRINTER and HOOKR on Robinhood Chain. Buy-to-post lead: FIH +37d, HOOKR +20d, AI +15d, QUOTRON +3d, SESTRI +1h43m, PRINTER -41m. He accumulated STONKBROKER from 19 Jul (day two), peaked 3,497,045 on 1 Aug and was fully out by 18 Aug without telling subscribers. His Solana wallet is still unidentified (he posted NOSIS FPfi9q1AixdUeWQVPFHJMJQ7a43S78dm6UZ4fzN4pump on 26 Aug)
- Method that cracked it, reusable for any paid X account: X keyword search DOES index a creator's subscriber-only posts for a paying subscriber, so `from:<handle> <term>` and `until:<date>` retrieve them going back years, while bare date-window enumeration (`since:X until:Y` with no keyword) is truncated to ~26 posts and unusable. Harvest by injecting a JS collector that writes to localStorage so it survives navigation between searches. Ghost posted no contract addresses at all before 14 Jun 2026, and his posting volume tracks the market: 9 posts in Jan rising to 71 in Aug
- Alert fleet change 27 Aug 2026: NEW BOT "Strong Trader Bot", service strongtrader.service at /opt/strongtrader (fleet is now 13). Built as a fork of rh-tracker/tracker.py. Watches the 5 Capture Board wallets on Robinhood Chain: 0x0428ef49 (50% win, 33:1), 0x7e3ba68c (59%, 4.8:1), 0x44df0854 fullturbo.eth (60%, 53:1), 0xde52c750 (71%, 6.0:1), 0xfa8e161b (57%, 138:1). Alert title = full address + win rate + payoff + positions, then what they did, then two DexScreener links (plain pair chart AND ?maker= filtered to that wallet's own fills on that coin) plus the tx and a wallet button. min_usd_buy and min_usd_sell both $1,000, 5s poll, alert_transfers false
- Three code changes in the strongtrader fork vs rhtracker: (1) new dex_pair_url() giving the unfiltered pair page alongside dex_maker_url(); (2) new config flag require_own_tx which, AFTER the quote-leg test but BEFORE the receipt is fetched, requires eth_getTransactionByHash(tx).from == wallet — this is what kills dust airdrops and other people's router txs that merely touch the wallet; the quote-leg path stays exempt because a matching opposite-sign quote leg is proof of the wallet's own trade however it was routed; (3) header rebuilt from a per-wallet "stats" object in config. Dry-run over 400,000 blocks (~11h) on live chain data fired 3 real alerts and classified many txs as "transfer (not-own-tx)", so the filter works
- quote_tokens for Robinhood Chain should be all three, not just WETH: 0x0bd7d308f8e1639fab988df18a8011f41eacad73 (WETH), 0x5fc5360d0400a0fd4f2af552add042d716f1d168 (USDG), 0x4eff9ccd650f6990132d1d738446256ca405f320 (WUSDG). rhtracker still only lists WETH and would misclassify USDG-paired swaps
- Cross-validation of the two screens: Ghost's real wallet 0x4123fFC81579e38dFFf74B75D8Bdf21c41e62237 independently surfaced in the Capture Board holder screen as a top-400 holder of SIX arc tokens (CASHCAT, AI, DELTA, INDEX, JUGGERNAUT, HOOKR — day-one on HOOKR) with a $295,955 book. It did not make the final six only because its position sizes are small next to the others. ghostwatch on the droplet already has it enabled and 0x96b13d47 disabled
- strongtrader.service is LIVE 27 Aug 2026 03:32 UTC. Bot @strongtraderfollower_bot ("Strong Trader Bot"), token 8811194495, chat 357464289. Alert has three link rows: DexScreener pair chart, DexScreener ?maker= filtered to that wallet, and a "his holdings" button to Blockscout /address/{wallet}?tab=tokens (holdings_url helper added)
- BUG AND FIX, remember this: running a dry-run harness that sets T.state["last_block"] and calls T.scan() WRITES that rewound pointer to the live state.json, because scan() ends with save_state(). On first service start strongtrader then backfilled ~350k blocks and alerted Ben on 26 Aug trades as if live. Fix applied: state pointer reset to head, seen cleared, and a new config flag max_catchup_blocks (set 9000 = ~15 min at 10 blocks/s) that clamps to head and logs "stale state" instead of replaying. Never dry-run against a service's own state file - copy it aside first
- Entry-proximity screen across the 5 Strong Trader wallets, 27 Aug 2026 (weighted-avg entry vs spot, only positions still held >$3k). STILL OPEN AT THEIR PRICE: NET on 0x7e3ba68c - $191,796 position built entirely on 26 Aug across 15 buys, ZERO sells, spot only +1% over its $1,145.69 avg entry, its single biggest new deployment; DTF on 0x7e3ba68c - bought 25 Aug at $0.00640, spot $0.00435, wallet is 32% UNDERWATER so entry is below theirs; MANCER on 0x7e3ba68c -17% (last buy 21 Aug); INDEX on 0xfa8e161b +16% (last buy 26 Aug, the only net buyer of the five); HOOKR on 0x7e3ba68c +28% (bought 24 Aug). TOO LATE: PONS +1095%, CASHCAT +3373%, HMM +3674%, INDEX on fullturbo +918%. WARNING not signal: TENDIES on 0xde52c750 is -29% below entry but that wallet is actively distributing it and rotating into BISCOTTI
- MAJOR CORRECTION 27 Aug 2026, found because Ben noticed the DexScreener maker link showed only SELLS for 0xde52c750's STONKBROKER and TENDIES. Two of the six Capture Board wallets did not BUY their headline positions, they RECEIVED them, so their cost basis, capture ratio, win rate and P&L were all fabricated by pricing an inbound transfer at market. Discriminator that settles it: a real DEX buy arrives from a contract (PoolManager 0x8366a39c, UniswapV3Pool, UniversalRouter 0x8876789976decbfcbbbe364623c63652db8c0904) with method 0x3593564c; a receipt arrives from an EOA with method 0xa9059cbb (plain transfer). Verify contract-ness with eth_getCode over RPC, NOT Blockscout is_contract - Blockscout intermittently returns {} on rate limit and a missing field reads as EOA, which produced a wrong classification mid-audit
- 0xde52c75041371b5e08d8a50b82c63734850628c1 is NOT a trader - DISABLED in strongtrader 27 Aug. Its whole book arrived in four plain transfers from ONE EOA, 0x9963597a9246b39b13330992f571f8378c18c262, between 01:13:46 and 01:14:24 on 5 Aug: CASHCAT 11,621,845 / STONKBROKER 31,734,095 / PONS 10,014,701 / TENDIES 18,222,063. Only its SELLS are real. Everything claimed for it earlier (+$2,517,251, 71% win, 100% capture, "bought the leaders with size and sat") is void. The source EOA 0x9963597a has 1,123 txs and 31,204 token transfers but only $99 left - it is a conduit that emptied itself
- 0x0428ef4923513f73d53fd6c28875574fc7ed0413 is MIXED, downgraded from #1. Its CASHCAT (the 24.96x "day-2 at $6.9M cap" story) was RECEIVED - 5,000,000 tokens in one transfer from EOA 0x9cc7474377da9c8fe6aa6bca8494a74bc1a88433 = wheelsmcfantini.eth (396 txs, $94,297 book: TENDIES $64,116, STACK $23,009, TYGR $6,798). Its PONS IS real: 3 DEX buys 15 Jul, 11.9x, 0.4% sold. Corrected verified record roughly 13 positions, 46% win, ~$292,911 deployed, ~+$251,985
- VERIFIED CLEAN, every position bought on-chain through routers/pools, no receipts: 0x7e3ba68c49561aae7c23c1d20fef0f1d7615a3ad (PONS 23 legs, CASHCAT 18, NET 15, MANCER 9, STACK 8, DTF 3, HOOKR 1), 0x44df085447dbebcf69c6675c3b8a795c7fdeb3f4 fullturbo.eth (CASHCAT 10, INDEX 6), 0xfa8e161b288386c38eced61ebdea0693de5184ff (HMM 29, PONS 8, INDEX 4). These three are the real answer to the early-and-hold screen; strongtrader now runs 4 wallets (these three plus the downgraded 0x0428ef49)
- Standing rule added from this: ANY wallet screen built on holder lists must audit the INBOUND legs before quoting a cost basis. A single large IN leg with no matching quote-asset leg is a receipt, not a buy, and holder-based screens cannot tell the difference