# Analyst 1, iterate run 22 (10 Oct): where the runners really are
Scripts: /opt/crew/work/analyst1/m*.py (research-box). Data: /opt/ppcopy e1x-runners.csv, r6usd/r6weth tapes, v4sender.db, poolout.db, review/misses.csv.
Focus rule: 8 of 10 iterations on the blind pool types (v4 no-hook WETH/USDG, unknown venue).

## Headline: the runner map is polluted three ways. After cleaning, the blind spot is USDG organic pools, not "v4 no-hook WETH".
Real runner = reached $1M AND >= $40k traded AND >= 50 swaps AND not a wash loop. Of 3,576 pools that touch $1M in e1x (21 Sep-9 Oct), only 614 pools (500 tokens) pass: 83% are dust/one-print jumps (2,415 trade < $2k in total).

| Pool type (token-level, clean) | real $1M | real $5M | who trades it | counterparties |
|---|---|---|---|---|
| v3 fee100 WETH (mc5 world) | 181 | 58 | live mc5 + paper | organic |
| v4 hooked 0x..a80 fee3000 (H3000) | 109 | 24 | h3k paper (13 trades) | **SCRIPTED: 9 buyers, 1 seller per pool** |
| v4 no-hook USDG, fee > 0 | 76 | 28 | nothing (USDG-EARLY sim only) | **organic: median 60 buyers / 62 sellers** |
| v4 no-hook WETH fee100 | 36 | 0 | busy60 etc. | scripted (6 buyers, top sender 99%) |
| v4 hooked USDG | 16 | 7 | nothing | |
| v4 no-hook native fee100 | 23 | 2 | | |
| SUSD/USDG fee 0 pairs | 9 | 9 | nothing | **WASH LOOPS** |

## 1. DUST RUNNERS (inversion). 
2,415 of 3,576 "$1M" pools trade < $2k. Prices come from 1-4 swaps in empty pools (v4 no-hook WETH custom fee: 0 real runners in 150+ pools; "v4 nohook fee10000 WETH 68 of 75 reach 5M" are all swaps <= 4 and volume $7). Discard as a hunt. These inflate every "runners per pool type" count.

## 2. LABEL BUG: the 14 "v4 no-hook WETH" misses of 7-9 Oct are H3000 hooked pools.
For 12 of the 14 (WOMBAG, MBLOOM, BMBFI, CRNKP, RROVR, PUPKEY, YLDCUB, BBDGR, STLZA, BPUFF, Clanker, ...) e1x shows two pools per token: a real hooked pool (hook ...a80, fee 3000, 8-26k swaps, $2-30M volume) and a decoy no-hook fee-3000 twin (2-150 swaps, < $100). The nightly review (and ppdata venue label) picked the decoy. So "extend 5-Min Cap to v4 no-hook WETH" was aimed at the wrong pool type; these runners are the H3000 family we already test (badly: 13 trades vs 109 runners). Needs a fix in the venue label: one label per TOKEN, taken from the pool with the most volume.

## 3. WASH LOOPS: "unknown venue" runners (PONS $385M, CASHCAT, AI, ORBIO, HOOKR, Muse, UBIK, V4 ...) are SUSD/USDG fee-0 pairs.
Same-block same-sender buy+sell loops: 76-96% of swaps in 9 of the top pools, 12-22 senders, volume $10M-$207M from ~300 swaps (single swaps of 750,000 USDG round-tripping). 17 of 129 fee<=3000 stable pools with >= 20 swaps are > 30% wash. They all "reach 5M" and none can be exited by an outsider at that price. Not a hunt. Unknown venue = a stable launcher with fee 0 or 1; ppfeeds should label it "stable-fee0-wash" and stop counting it as a miss.

## 4. SCRIPTED RAMP = H3000 a80 (new lens: how many different wallets sell?)
119 a80 pools with tape: ~10,800 swaps each, buyers median 9, sellers median 1 (100% of pools <= 2 sellers), top sender = 99% of buys AND 99% of sells. Price ramps 2.2x at +1 h minimum, median max 84x, 0 pulls ever. One wallet makes all the volume and the only sell. In 119 pools no outsider has ever sold. Replay 4.6x tp5 is a price path that nobody else has realised. This is exactly Ben's pending decision (#221, $5 real sell): my number makes it the highest-value $5 we can spend (109 runners at stake). Control: USDG no-hook runners 60 buyers/62 sellers; v3 fee100 not measured here.
Corollary: in the v4 no-hook WETH fee100 "36 real runners" top sender is also 99% (scripted, 0 reach 5M).

## 5. LAUNCH FEE AS METADATA (new lens: pool parameters as a label)
USDG/SUSD v4 pools carry the fee the launcher chose (13,552 pools). Real 5M runners by fee: fee 0: 21 (wash), 1-3000: 11, 3000-10000: 11, 10000-50000: 42; fee 50k-100k: 1 of 1,387; fee >= 100k: **0 of 7,794 pools** (plus 436 pools with fee > 2M: 9, all dynamic-fee). Rule at birth: skip fee in (50k, 2M]. 7,794 pools removed with zero real runners lost.
In the USDG-EARLY sim (a1-usdg-trades.csv) the fee>100k cells (1.73 mean) are phantom gains; fee <= 50k: n=762 mean 1.52, halves 1.49/1.54, w/o top5 1.21.

## 6. USDG-EARLY-B (READY FOR PAPER): all-buys start
Universe: USDG/SUSD v4 pool, fee 1..50,000 (wash fee 0 excluded), first-3-swap median cap < $300k. Feature known at entry: number of sells among the first 4 swaps. Entry = after swap 4; stop 0.5x; trail 0.6x of high after 2x; fee per side = pool fee (min 1%).
| first 4 swaps | n | mean x | median | halves | w/o top5 |
|---|---|---|---|---|---|
| all pools fee<=50k | 699 | 1.44 | 0.96 | 1.46/1.42 | 1.17 |
| >= 2 buys | 507 | 1.66 | 0.99 | 1.67/1.65 | 1.29 |
| >= 3 buys | 392 | 1.81 | 0.99 | 1.77/1.85 | 1.34 |
| 4 buys (0 sells) | 259 | 2.21 | 0.99 | 2.19/2.23 | 1.51 |
4-buys cell: win 38%, 8 trades >= 10x, top5 = 33% of profit; quarters 2.82/1.53/1.33/3.18. With fee > 0 only (excludes wash fee 0): n=157, 1.99, w/o top5 1.52. Hooked n=126: 3.0 (w/o top5 1.58); no-hook n=133: 1.46 (1.14). Lottery-shaped (median 0.99). Not modelled: slippage on thin books (use reserve >= 1 ETH-equivalent check), entry latency, USDG hook/tax sellability: the paper test must do a quote/sell check per entry. Distinct-sender count in the first swaps carries no signal (nsend>=4 0.84 w/o top5).

## 7. REGIME (BTC/ETH) vs clean runner rate: DISCARD
19 days, clean pools per day vs share reaching $1M: Spearman vs BTC day return 0.14 (p=0.57), vs ETH 0.17, vs previous-day BTC 0.18, previous-day count 0.05. BTC dump days (23 Sep, 7-8 Oct: -2.7/-2.6/-2.9%) had 0.148/0.141/0.136 vs mean 0.16. Daily real $1M runners born: 21-63 (mean 30), 5M: 5-18. Ben's 15-loss streak is variance, not BTC. Counts only drop with chain-wide launches (n/n_prev rho 0.55).

## 8. Fee value granularity within USDG (DISCARD): fee 50000 (n=405) 1.7% reach 5M, 30000 2.2%, 20000 3.3%, 10000 1.9%, 9000/2500/3000/500/75 0. Only "sub-3000 tiers = junk" (0 of 400+).
## 9. Twin decoy as signal (DISCARD): a80 pools 130, decoy no-hook fee-3000 twins 132, 1:1, no information.
## 10. Data requests: see below.

## Requests
- ppdata (posted): add per-pool columns to e1x-runners.csv and refresh to today; one label per token.
- ppfeeds (posted): a "scripted-pool" flag in the v4 feed (top-sender share of buys in first 60 s >= 0.9 and distinct buyers <= 12) and a wash flag for stable pools.