RSI(2) control check (not part of the frozen protocol). Run 2026-10-07T18:56:42.259Z

=== A (original exits, long + short) ===  real: n=674 net 25243 avg 37.5 $/trade  (engine copy reproduces the published trades: yes)
  C0 frozen replica (any bar, random side): trades mean 651 (real 674); random median -25.9, 5th -82.7, 95th 32.0 $/trade; real 37.5 -> percentile 97.2
  C1 trend-matched (long above SMA200, short below): trades mean 653 (real 674); random median -16.9, 5th -74.9, 95th 34.0 $/trade; real 37.5 -> percentile 95.8
  C2 trend + dip state (long: above SMA200 and below SMA5; short: mirror): trades mean 665 (real 674); random median -3.1, 5th -63.2, 95th 59.8 $/trade; real 37.5 -> percentile 85.4

=== B (original exits, long only) ===  real: n=412 net 26102 avg 63.4 $/trade  (engine copy reproduces the published trades: yes)
  C0 frozen replica (any bar, random side): trades mean 390 (real 412); random median -8.7, 5th -80.5, 95th 74.9 $/trade; real 63.4 -> percentile 91.6
  C1 trend-matched (long above SMA200): trades mean 407 (real 412); random median -4.9, 5th -54.3, 95th 49.7 $/trade; real 63.4 -> percentile 98.6
  C2 trend + dip state (long: above SMA200 and below SMA5): trades mean 400 (real 412); random median 18.0, 5th -40.7, 95th 73.4 $/trade; real 63.4 -> percentile 90.2

=== A2 (Round 2 stop and target, long + short) ===  real: n=746 net 3625 avg 4.9 $/trade  (engine copy reproduces the published trades: yes)
  C0 frozen replica (any bar, random side): trades mean 734 (real 746); random median -37.1, 5th -125.4, 95th 58.3 $/trade; real 4.9 -> percentile 74.2
  C1 trend-matched (long above SMA200, short below): trades mean 732 (real 746); random median -36.3, 5th -122.2, 95th 46.5 $/trade; real 4.9 -> percentile 81.4
  C2 trend + dip state (long: above SMA200 and below SMA5; short: mirror): trades mean 755 (real 746); random median -41.7, 5th -113.6, 95th 32.1 $/trade; real 4.9 -> percentile 85.0

=== B2 (Round 2 stop and target, long only) ===  real: n=453 net 16239 avg 35.8 $/trade  (engine copy reproduces the published trades: yes)
  C0 frozen replica (any bar, random side): trades mean 436 (real 453); random median 0.6, 5th -98.6, 95th 111.1 $/trade; real 35.8 -> percentile 71.4
  C1 trend-matched (long above SMA200): trades mean 441 (real 453); random median -10.0, 5th -90.0, 95th 73.8 $/trade; real 35.8 -> percentile 82.6
  C2 trend + dip state (long: above SMA200 and below SMA5): trades mean 439 (real 453); random median -38.8, 5th -109.3, 95th 36.3 $/trade; real 35.8 -> percentile 94.8