# Miss review 15 Oct 2026 (analyst1, board #2578)
19 runners not caught. Source: review/misses.csv + gold_idx (on-chain index, main:/opt/pricepredict-data/gold_idx.json).

## Verdict per runner
| runner | ATH | kind | why missed | any paper test? |
|---|---|---|---|---|
| 0x5199ea | 26B (fake price) | v4 native fee883131 jump, cap_first 366k | seen, cap 5.1M at 5 min; ATH is a price artefact, not tradable | no |
| Dacity | 21.7M | v4 native organic (depth 27.7k, cap_first 6.8k) | **NOT SEEN** by mc5 (venue pool_uniswap_v4). The one real big runner. | no |
| 0xfd6117 0xfb73a9 | 6.9M, 6.2M | kit:other | seen, "in band but skip" (no radar / thin v3) | no |
| 0x9bb9f9, 0xe01d7e | 5.7M, 1.0M | kit:other v3 thin/organic | seen, cap under 250k at 5 min (rule) | no |
| 0x7dbc3d | 5.1M | kit:other v3 jump | cap 5.1M at 5 min, rule | no |
| **8 x kit B** (0x64b08e 4.0M, 0x4cd515 3.3M, 0x94653c 3.3M, 0x324dbd 2.8M, 0x39aaa9 2.6M, 0xfe8fc6 2.5M, 0x1b7669 2.4M, 0x434beb 2.0M) | 2-4M | v3 WETH fee100 kit code 89885b75fd29 | seen, cap 1.5-2.3M at 5 min, over the $1M ceiling (rule: too late) | no |
| AGEM, WIN | 1.4M, 1.2M | standard_reserve / v4 tokq hooked | **NOT SEEN** (venue not watched by mc5); hooked tokq is already a radar class | no |

Technical failures: 0. Not seen: 3 (Dacity, AGEM, WIN). Rule: 16.

## The pattern: kit B is a runner factory, and we only look at it 4 minutes too late
Kit B = v3 pool, WETH, fee 100, creator code 89885b75fd29 (gold_idx). Since 1 Oct: **478 pools, 131 (27%) reached $1M, 41 reached $3M**, 301 distinct creators (not one operator, cannot blacklist or follow a wallet). Daily hit rate swings: 3 Oct 2/40, 11 Oct 4/72, but **13 Oct 19/21, 15 Oct 15/18, 14 Oct 11/27**. A wave state.
- Time from first swap to $1M cross: deciles 0 / 47 / 68 / 112 / 424 s (median ~110 s, 30% inside 70 s). 273/478 touch 300k; 131/273 (48%) of those go on to 1M.
- Liquidity pull comes late: after the 1M cross median ~35 min (p10 9.5 min, 75 pools measured). So an exit at the 1M cross or at 2x of entry leaves a long window; the pull is not the exit constraint, the entry is.
- mc5 enters at 300 s, when the cap is already 1.5-2.3M, over its ceiling, so it correctly skips (V3-R300 on SEALS lost 65% buying such a spike). The money is in entering at the 100-300k cross (inside the first 0-60 s) and selling at the 1M cross: gross 3.3-10x on 48% of 300k touchers.

## Break-even maths (not yet money)
Entry at 300k cross, exit at 1M cross: win 3.3x (before fees/slippage; real fill of a 3x trade at $10 ~ 2.8x). P(win)=48% over all days. Break-even with loser leg L=0.4: P >= 28%, with L=0 (rug): P >= 36%. On wave days P is 70-90%; on dead days (3 Oct, 11 Oct) it is 6-20% and loses. So the **wave gate matters more than the entry rule**: take kit B entries only when the last 5 completed kit B pools (4 h window) had >=2 reach 1M. Unknown, and the reason this is not yet a test-ready rule: the loser leg (what a 300k toucher does when it fails, do bots sell into it, outside-wallet sellability) needs a per-swap tape. gold_idx has crossing times only; v4tape does not cover v3. Need that tape before testmgr.

## Proposed one change
Early kit-B lane: watch every new v3 WETH fee100 pool with code 89885b75fd29 from the first swap; buy at the 300k cross if the wave gate is open; exit at 1M cross (or -50% stop / 5 min no progress). Catches 8 of 8 kit-B misses of 15 Oct on paper (all crossed 300k at birth, 1M 44-110 s later) and 131 of 131 historic kit-B runners that crossed 300k early, but the losers (142 touchers that fail) are unsized.

## Data request (ppdata)
Per-swap tape (ts, block, amount in/out, price/cap, sender, tx.from) for all gold_idx pools with code 89885b75fd29, 1-15 Oct (478 pools; first 30 min of each, or until pull), to size the loser leg and run the wave gate on 30 days.

## Not-seen fixes
- Dacity (21.7M, v4 native organic, depth 27.7k): mc5 does not watch v4 native organic pools. Biggest value of the day, route to ppfeeds [miss-watch]: add v4 native organic to the radar/feed and count how many runners in 7 days were in that class.
- AGEM/WIN (standard_reserve, v4 tokq hooked): known hooked tokq class, already in radar; not a trade venue for mc5 (ledger hooked family).