# analyst3: 10 Oct misses (22 runners) -> patterns and testable ideas (#1707)
Data: review/misses.csv (10 Oct), radar.jsonl crossings, pool_paths.jsonl (13 Sep-11 Oct). Scripts m1-m3.py (work/analyst3).

## Where the 22 sit
| group | runners | what it is |
|---|---|---|
| v3 fee100 WETH, seed (first print) <12k, cap ~0.6-2M at +1-5 min | 8 (VORTA 20M, hoodcat 17M + 2.9M, PAWS 6.6M + 1.0M, DAEMON, SIB, roflmao-ish) | born-low seed, jump to ~1M inside 1 min, then run. mc5 band (250k-1M at 5 min) excludes the ones above 1M (hoodcat 2.2M, SIB 1.1M); VORTA 957k WAS in band. |
| H3000 family (fee 3000 hooked WETH), born ~30k, 300k at age 10-40 min | 6 (FMRGN? see below, MGCMP, HBLINU, BRML, CHNCHL, ADWYR) | review labels these "pp:v4/no hook/WETH"; radar says v4:weth:hook. Already A2-H3000 (+60 s) and a2-h3000-late300k (92% reach 1M). Technical mislabel in review + fml2 skip_sellquote on all 5. |
| USDG/ETH/other | 6 (QQQ, AAPL, Paper, hippo, OP, ...) | pinned/decoy risk, known (see a3-iter-o). |

Radar check, H3000 born-small (cap_first<100k) crossing 300k: 10 of 15 reached 1M (all 10 Oct 05:00 - 11 Oct 04:28 UTC, 1M at 19-152 min after, depth 150-180k at 1M). Consistent with A2 late300k: nothing new, but the review's venue label hides them. -> reviewer/manager: fix label (hook != 0x000 => hooked).

## New: seed-small v3 fee100 above the mc5 band (hunt: entry +5 min at 1-3M cap, target 5M+)
Base: v3 WETH unhooked fee100, pool_paths 7,460 pools. Crude path sim (cap_5m entry, checkpoints 15/30/60 min + peak, tp5 sl.5, 3%/side; stop only seen at checkpoints = optimistic):
| cap at 5 min | seed<12k | seed>=12k |
|---|---|---|
| 250-500k | n127 mean 1.85 hit5 24% | n47 1.06 hit5 4% |
| 500k-1M | n85 2.27 hit5 33% (halves 2.24/2.30) | n44 0.96 hit5 0% |
| 1-3M | n27 2.53 hit5 41% (halves 2.76/2.32) | n20 0.95 hit5 0% |
Seed split is the whole story (matches A3-SEED). Above 1M the band rule drops 27 pools in 18 days of which 11 reached 5x-ish; 13 of 27 are 24-30 Sep, 3 in the last 3 days (hoodcat 2x, SIB) so rate ~1-2/day and regime dependent. Depth at 5 min median $62k: size 0.04 ETH ok, exits into 5M are the open point (depth is a fixed curve).
Caveat: peak/checkpoint sim, not a fill sim; ordering of peak vs stop unknown.

Ideas:
1. A3-SEEDHI (testmgr forward arm): v3 fee100 unhooked WETH, first-print cap <12k, cap at +300 s 1-3M, depth >=40k, enter +300 s, tp5 / tp10 arm, stop 0.5, 4 h. Hunt: entry 1-3M, target 5M+. Kill: mean <1.3 over 20.
2. Review label: hooked v4 shown as "no hook" (reviewer).