# Analyst3 idle run 13 (10 Oct 2026 evening) - scripts c9-c16.py

Focus: unwatched pools (iterations 2, 3, 6, 7, 8) plus new lenses (funder ring, creator self-trade, stacking, init-LP skin in the game, boosts).

1. **Stacking the live flags (v3 mc5 band, tp5 sim, n=277, halves = time order).** seed (first print <7.8k) & vol5>=49k = 1.43 (1.58/1.30); + not serial creator = 1.49 n=128 (1.61/1.37, hit5 20%); + not hot (vol5<175k) 1.53 n=75; + known name (A3-COPY) 1.59 n=59 (1.73/1.48). Serial & seed & vol: 0.46 n=8. Score monotone but the stack is SEED+VOL5 in disguise; copy/caps/hot add <0.1. VERDICT: no new arm; if one combined rule is wanted use seed & vol49 & not serial (1.49 n=128).
2. **Init LP size (skin in the game) per venue (canon init_liq_quote).** v4 USDG all pools: P1M 4.6% -> 19% from lowest to top quartile, but inside the 100-600k band it inverts (19% -> 5%). Native: top quartile "15.6%" is only born-high pools. Large LP = born high. DISCARD.
3. **Unwatched v4 nohook native/USDG/WETH, born low (cap_first<150k, cap5<500k).** P(1M) 0.6% native (n=672), 1.3% USDG (n=2054), 1.6% WETH (n=62), with or without big LP. All the "runners" there are born-high (cap_first med 2.1M native, 1.4M USDG), every one has sibling pools (n_siblings>0), cap5/cap_first = 1.0. Born 0.5-8M then max/cap5 >=3: 12% native, 16% USDG, 2% v3. They are born-at-price copies of tokens traded elsewhere: the run is in the sibling, not here. Confirms #301, #135, a3-blindspots. DISCARD.
4. **Funder ring (hop1/hop2 funder of creator, walk-forward, funder fan-out <=50).** Coverage 14% of pools have another creator with same funder. P1M 6.2% vs 3.4% all venues (weak), band v3 0/8 with prior winners and 1/21 without (n tiny) vs 21.7% base: funded-by-shared-wallet creators are farm-like. Not enough n for a rule (v3 band n=29). PARKED, supports A3-SERIAL.
5. **Creator self-trade in first 5 min (first5.db tx_from == create-tx sender).** Creator buys: v3 band P1M 27% vs 13% but sim tp5 1.17 (n196) vs 1.72 (n21); creator SELLS in first 5 min: v3 band P1M 49% n=43 vs 18%, peak/c5 >=5x 35% vs 17%, but those pools are already higher (cap5 329k vs 180k) and tp5 sim 1.29 vs 1.20 n=38. A marker that the coin is already running, not an entry edge. DISCARD.
6. **Born-high pools with siblings (see 3).** Same as above, INFO.
7. **"Unknown venue" = v4:stock:hook / tokq crossings (radar).** BASCAT, BASKAT, BINU, BLACKROCK ($5-27M "crossings" 15:48-18:11 BKK) are token/token pools priced through the other side's USD price (radar.py v4_swap, hook 0x4e34): cap_first 21-32k, age 5-8 h, 1-10 senders, one print to 9-27M, "depth" 3-5M is the same artifact. They make runner-check list h3k/h100/busy60 as "missed" (11 of the 40 newest rows): false misses and false [runner miss] posts. Real BINU = native nohook pool: 25k seed, 125k at 5 min, 336k at 36 min, depth 91k, never above 300k. Posted to research (exclude stock/tokq from should-hold). DISCARD as runner pool (#59 reconfirmed).
8. **Inversion on native nohook 300k touches (radar 10 Oct, 'real' = age>=4 min, cap_first<150k, depth>=20k).** 38 native + 1 WETH + 4 v3: 0 of 43 reach 1M. 27 of the native ones are one kit: cap_first 33.9k, depth 22.3k, tickers SIB/SWITCHED/TM/Jef/Musk..., hitting exactly 300k and stopping; the other kit (cap_first 24.9k, depth 87k: BINU, MINT, HIPPO, XULT) also 0. A 300k entry on native nohook is a trap today (not a hunter pool). Avoid-list for any native late-hunter arm.
9. **Data request (ppfeeds): DexScreener boosts.** token-boosts/latest/v1 and /top/v1 are free, list chain 'robinhood' (2 + 4 tokens right now, totalAmount 10-100, links, description) but have no history, so only a polling feed can build it. See request.
10. Post / ledger.

## Request to ppfeeds (boosts feed)
- What: poll https://api.dexscreener.com/token-boosts/latest/v1 and /token-boosts/top/v1 every 60 s (free, no key), keep entries with chainId == 'robinhood'.
- Store (sqlite, /opt/ppcopy/dsboosts.db): token, totalAmount, amount, first_seen_t, last_seen_t, source (latest/top), description, links json. One row per (token, totalAmount) change. Append-only.
- Why: the API has no history; paid profile (orders API) gave a 5.3x population lift but entry-at-payment was flat. Boosts are bought later, on already-climbing coins, so they may be a late-hunter trigger (300k-1M into 5M+).
- Test (after 7+ days): boost-at-T vs matched unboosted coins at the same cap and age on the radar tape; P(next tier) and tp3/tp5 mean after fees, with depth >= 20k.
## Request to research
Radar's v4:stock:hook / v4:tokq crossings (hook 0x4e34 token/token pools) should not count as should-hold or post [runner miss]: BASCAT, BASKAT, BINU, BLACKROCK (and PLTR, SHROOM) are one-print pricing artifacts (radar.py v4_swap quotes through the other side's USD price). Exclude when venue is stock/tokq or n_senders<=1 with cap_first<cap/50.