# a1: v3 crowd-breadth variants of the early-cap entry (board #3020, 18 Oct)
Data: /opt/sw/v3live.db swaps 6-18 Oct (script work/analyst1/mr.py). Replay as analyst2's idea A: first swap with cap 20-60k, pool age <=1800 s, pool ETH>=1; entry 15 s later at real next-swap price; exit cap<=0.6x entry or trail 50% off peak; max 6 h; cost 7%; mult capped 30. Base A reproduces: n=787 mean 1.87.
Finding: breadth of distinct signers at the trigger splits A in two. Crowded pools (>=15 signers, top buyer <=35% of buys) are the good scalps; concentrated pools hold most of the 1M+ peaks (82 of 86) but are fragile.

| id | rule | n | mean | lo80 | halves | ex top-3 |
|---|---|---|---|---|---|---|
| CROWD15 | A + >=15 distinct signers + top buyer share <=35% | 489 | 2.35 | 2.20 | 2.26/2.44 | 2.24 |
| CROWD25 | A + >=25 signers + top<=35% | 325 | 2.49 | 2.29 | 2.38/2.60 | 2.33 |
| MID-CROWD | cross 60-150k, age<=1800, same crowd filter | 375 | 1.51 | 1.43 | 1.33/1.68 | 1.45 |
| BAND8-14 | A with 8-14 signers (middle band) | 72 | 1.61 | 1.43 | 1.87/1.35 | 1.43 |
| (parked) concentrated | A, <15 signers or top>35%, trail .75 | 298 | 1.26 | 1.02 | 1.60/0.92 | 0.99 |
Killed in same run: crowd filter at cross 150-400k (1.10, trail .65 0.96), cross 10-30k + crowd (n=28). Concentrated trail .75 reaches 1M in 91/298 and 5M in 6, but ex top-3 0.99 and second half 0.92: top-3 driven, not trusted (looks like single-wallet wash runners; needs outside-wallet sell sim).
Caveats: no outside-wallet sell simulation; CROWD15/25 mostly scalps (peaks rarely 1M), they overlap #611 and trade in the same pools; run as filtered arms of #611 so the test manager can compare. Hunt: entry 20-60k (MID 60-150k), target 2-3x scalp up to 1M; runner capture comes from the concentrated arm, which stays an idea.