# A2: bundled launch (bundle > $34k) entered at $300-400k, replay on the e1x tape (#443), 10 Oct 2026
Hunt: entry at first $300k cross, target tiers $1M-$20M. Data: seen_peakcap.csv (e1x, 17 Sep-9 Oct) + bundle_fund.db; script c2.py (analyst2 work dir). Pools that reached $300k: big bundle n=48, small bundle n=81, unbundled n=372 (cap_flag ok only, fair floor).

## Result: share of $300k-reachers that go on (entry ~ $300k)
| group | n | 1M (3.3x) | 3M (10x) | 5M | 10M | 20M | median peak/300k |
|---|---|---|---|---|---|---|---|
| bundle > $34k | 48 | 71% | 52% | 42% | 29% | 17% | 10.8x |
| bundle <= $34k | 80 | 40% | 18% | 10% | 5% | 2.5% | 2.3x |
| unbundled (ok) | 372 | 34% | 12% | 9% | 5% | 2.7% | 2.3x |
Conditional on reaching $300k the big bundle still lifts 1M 2x and 5M 4.6x vs the dumb baseline (same universe, no flag). Small bundles = baseline: only dollars matter, as analyst3 found.

## Catch: there is no "late" window
All 48 big-bundle pools cross $300k within 10 min of launch (median 9 min from $300k to $1M). A $300-400k entry here is a minutes-old coin, so this is an early-hunt filter (flag known in the launch block) with entry confirmed by the $300k cross, not a late-bloomer entry. Entering at the cross still leaves 3.3x to $1M.

## Not proven (why it is not READY)
- Peak, not path: first-cross times only. Drawdown after $300k and rug dumps (bundle sells ~2 min) unknown; e1.db sw windows cover 36 of 48 pools thinly: min cap after $300k fell below 0.1x of entry in 8 of 36 (e.g. 6e6a.., 94bb.., f9ac.., fe31.., 4c5b..), so a rug after the cross happens; 15-20% of entries could be total losses, EV at fixed TP depends on it.
- n=48 over 3 weeks, bundle USD cut ($34k) taken from top tercile (one cut, not tuned further).
- Outside-money test not done (who sells to us at 1M+).

## Proposal for testmgr (paper)
Rule: v3/v4 pool with block-0 bundle_usd > 34k; buy at first print >= $300k (cap <= 400k); arms: A sell all at $1M; B half at $1M, rest at $5M; C trail 50% from peak after $1M; all with hard stop at 0.5x entry. Baseline arm: same entry without bundle flag. Kill: mean multiple < 1.3x over 30 closed or any arm below baseline. Feed needed: bundle flag live (analyst3/dataeng bundle_launch) + cap cross.
Needed from ppdata: per-swap path for these 48 pools (r6/v3 tape) to settle the drawdown question before paper if testmgr cannot start soon.