# A3-BREADTH forward paper test (Test Manager) - 11 Oct 08:40 UTC
Hunt: v3 mc5 band (cap 250k-1M at +300 s), entry first swap >= +306 s, tp5 sl.5 4 h, target 5M+; filter = breadth: distinct tx.from buyers in first 300 s >= 30 (A3 #928, fw/a3-iter-10oct-p.md).
Arms: A = nfrom>=30 | B = nfrom>=30 AND seed<7.8k AND vol5>=49k | C = every band pool (baseline). 7% slippage, pull=0.
- tx.from resolved by us via free RPC (v3live has no tx.from): 0 looked up this run, 0 pools still pending (backfill resumes each run, newest first).
- Replay since 6 Oct (in-sample, threshold fitted by A3): 70 band entries resolved
  - A: n 45 closed, mean 0.99x, median 0.44x, win 6, hit5 6, stop 38, pull 0, time 1
  - B: n 28 closed, mean 1.18x, median 0.45x, win 5, hit5 5, stop 22, pull 0, time 1
  - C: n 70 closed, mean 0.90x, median 0.45x, win 7, hit5 7, stop 48, pull 2, time 13
- FORWARD only (pools created from 10 Oct 00:00 UTC): 12 band entries
  - A: n 6 closed, mean 0.51x, median 0.45x, win 0, hit5 0, stop 5, pull 0, time 1
  - B: n 2 closed, mean 0.65x, median 0.85x, win 0, hit5 0, stop 1, pull 0, time 1
  - C: n 12 closed, mean 0.52x, median 0.45x, win 0, hit5 0, stop 9, pull 0, time 3
- Skipped: {'cap300 outside 250k-1M': 1408, 'no swaps by 300 s': 70, 'no swap after +306 s yet': 6}
- Kill (A3 sim A 1.57 vs .83; expect lower): forward arm A mean < 1.1 over 30 closed. Retire CAPS/COPY arms (A3: noise).
- Limits: mark-to-print exits; entry fill depth not audited. Only Ben switches live (D-030).