fix(rl): surfer-scaffold v5.2 — drop wr_excess clamp, full scaffold when novice

alpha-rl-mjsmr (4a4953b01 step 99-249, 2026-06-02): v5.1 with wr_ema=0.248
(below break_even 0.30) produced scaffold_weight = 0.500, not the expected
1.0. Math: `max(0, wr_ema - break_even) = 0` → sigmoid(0) = 0.5 → competence
= 0.5 → w_competence = 0.5. So a novice agent below break-even got
HALF-scaffold instead of full scaffold (entropy dropped to 1.27 at step 249
because of the early directional signal, then ramped back up to 1.96 once
w_decay took over — unstable).

Fix: drop the `max(0, ...)` clamp. Use signed `wr_distance = wr_ema -
break_even` so:

  wr=0.20 → sigmoid(30·-0.10) ≈ 0.05 → w_competence ≈ 0.95 (full scaffold ✓)
  wr=0.30 → sigmoid(0)        = 0.50  → w_competence = 0.50 (half at BE)
  wr=0.40 → sigmoid(30·+0.10) ≈ 0.95 → w_competence ≈ 0.05 (pure pnl ✓)

This is the proper symmetric fade around break_even; the asymmetry was a
v5 original-design error. Combined with v5.1 `w_decay = max(0, 2·frac-1)`,
the scaffold should now:
  - hold near 1.0 while wr < break_even (novice)
  - smoothly fade through 0.5 as wr crosses break_even
  - drop toward 0 as wr settles above break_even AND PH decay subsides

Local invariant test:
  reward_alignment_surfer_scaffold_invariants OK: 251 rows validated
  train[0] bootstrap = 0.966 (unchanged)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

crates/ml-alpha/cuda/rl_surfer_scaffold_controller.cu

@@ -19,19 +19,26 @@
 //   RL_SURFER_SCAFFOLD_WEIGHT_INDEX (753)  — w ∈ [0,1] consumed by
 //                                            rl_fused_reward_pipeline.cu Phase 5
 //
-// Math:
-//   competence = confidence(n_trades) × sigmoid(k_sharp × max(0, wr_ema − break_even))
+// Math (v5.2):
+//   competence   = confidence(n_trades) × sigmoid(k_sharp × (wr_ema − break_even))
 //   w_competence = 1 − competence
-//   w_decay      = max(0, 2 × frac_alerted − 1)   // FIX v5.1: only fires above 50%
+//   w_decay      = max(0, 2 × frac_alerted − 1)   // v5.1: fires above 50% baseline
 //   w            = clamp(max(w_competence, w_decay), 0, 1)
 //
-// At training start (n_trades=0, wr_ema=0): competence ≈ 0, w_competence ≈ 1 → w=1 (full scaffold)
-// After warmup and wr_ema crossing break-even: competence → 1, w → 0 (pure pnl)
-// If edge-decay PH alerts on >50% of batches: w_decay re-engages PROPORTIONAL to
-// excess above 50% baseline (alpha-rl-zf6s5 step 5719 verdict 2026-06-02 showed
-// train-time PH inherently alerts on 75-95% of batches, so the v5 original
-// `min(1, 2×frac)` pinned w=1 forever even when wr crossed break-even; the new
-// `max(0, 2×frac − 1)` only fires above 50% and reaches 1.0 only at 100% alert).
+// v5.2 fix (alpha-rl-mjsmr step 99-249 verdict 2026-06-02): dropped the
+// `max(0, …)` clamp on `wr_excess`. With the clamp, wr<break_even mapped to
+// sigmoid(0)=0.5 (half-scaffold at novice agent — wrong). Now wr=0.20 maps
+// to sigmoid(−3)≈0.05 ⇒ competence≈0.05 ⇒ w_competence≈0.95 (full scaffold,
+// correct novice-bias). Symmetric around break_even — competence saturates
+// fast both directions for k_sharp=30.
+//
+// Behavioral table:
+//   wr_ema     wr-bk    sigmoid    competence (n_trades≫warmup)    w_competence
+//   0.20      -0.10    0.047       0.05                            0.95   (full scaffold)
+//   0.27      -0.03    0.29        0.29                            0.71
+//   0.30       0.00    0.50        0.50                            0.50   (half at break-even)
+//   0.33      +0.03    0.71        0.71                            0.29
+//   0.40      +0.10    0.953       0.95                            0.05   (pure pnl)
 //
 // Per `feedback_no_atomicadd`: single-thread launch (1×1×1), no atomic ops.
 // Per `feedback_cpu_is_read_only`: pure device-side.
@@ -71,11 +78,12 @@ extern "C" __global__ void rl_surfer_scaffold_controller(float* isv) {
     const float warmup_scale = fmaxf(warmup_n * 0.3f, 1.0f);  // floor to avoid div-by-zero
     const float confidence   = sigmoidf((n_trades - warmup_n) / warmup_scale);
 
-    // Win-rate excess above break-even, gated nonneg.
-    const float wr_excess = fmaxf(0.0f, wr_ema - break_even);
+    // Win-rate distance from break-even (signed — v5.2 drops the max(0,...) clamp
+    // so wr<break_even properly maps to low competence ⇒ full scaffold).
+    const float wr_distance = wr_ema - break_even;
 
     // Competence: agent must have BOTH trade-count confidence AND wr above break-even.
-    const float competence = confidence * sigmoidf(k_sharp * wr_excess);
+    const float competence = confidence * sigmoidf(k_sharp * wr_distance);
 
     // Inverse: scaffold weight from competence.
     const float w_competence = 1.0f - competence;