# A3-WIDE forward paper test (Test Manager) - 12 Oct 02:10 UTC
Hunt: mc5-style v3 band entries (cap 200k-1.5M at +300 s, >=20 swaps, buy first swap >= +306 s), target 10x-20x (3M+ tier). Rule from fw/a3-iter-10oct-f.md item 3 (#572). 7% slippage (mult*0.93); pull=0; 4 h.
Arms: A tp10 sl.3 | B tp20 sl.3 | C tp5 sl.5 (baseline).
- Replay since 6 Oct (in-sample): 82 entries
  - A: n 82 closed, mean 1.19x, median 0.27x, win 9, tp 8, stop 62, pull 8, time 4
  - B: n 82 closed, mean 1.62x, median 0.27x, win 8, tp 3, stop 63, pull 8, time 8
  - C: n 82 closed, mean 0.86x, median 0.45x, win 9, tp 9, stop 67, pull 5, time 1
- FORWARD only (pools created from 10 Oct 00:00 UTC): 26 entries
  - A: n 26 closed, mean 0.60x, median 0.27x, win 1, tp 1, stop 21, pull 3, time 1
  - B: n 26 closed, mean 0.75x, median 0.27x, win 1, tp 0, stop 21, pull 3, time 2
  - C: n 26 closed, mean 0.55x, median 0.45x, win 1, tp 1, stop 21, pull 3, time 1
- Replay vs forward, n / mean (80% CI), trades >=10x:
  | arm | replay | forward | >=10x replay/fwd |
  |---|---|---|---|
  | A | 82 / 1.19 (0.81-1.58) | 26 / 0.60 (0.15-1.05) | 0/0 |
  | B | 82 / 1.62 (0.99-2.24) | 26 / 0.75 (0.11-1.38) | 7/1 |
  | C | 82 / 0.86 (0.67-1.05) | 26 / 0.55 (0.33-0.76) | 0/0 |
- Verdict basis is FORWARD only (auditor 11 Oct): replay mean of B rests on a few >=10x fills. Re-audit at forward n>=30 per arm. Depth walk at 10x/20x still open.
- Skipped: {'cap300 outside 200k-1.5M': 798, 'swaps<20 by 300 s': 846, 'no swap after +306 s yet': 1}
- Kill (my rule): after 100 forward closed entries arm A mean < 1.3. A3 sim: A 1.67, B 2.15, C 1.28 (tp hits rare: 16 for tp20).
- Limits: mark-to-print exit, NO real sell quote at 10x (pool depth unproven; A3 asks for it). Fill depth beyond 7% slip not modelled. Only Ben switches live (D-030).