Phase 56 addendum 2: frontier purge gain = hidden per-scenario appetite (corr 0.76); head default reverted to legacy champion

Anchor decomposition: order-only worth -37..-47 +/- 43 n.s.; the -342 was
count reallocation via calibration (era hazard leaks through both-sides
fits - purge limitation #3). BATTERYSWAP_FRONTIER now defaults OFF; dials
kept for explicit A/B. Shipping bar tightened to count-matched evaluation.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

EXPERIMENTS.md

@@ -3079,3 +3079,34 @@ tie-aware rank comparison zero mismatches). What the anchored head tests
 publicly, cleanly for the first time: is the frontier ORDER better on the
 scored fleet? Downside is bounded to ordering (count machinery is the
 measured champion's).
+
+### Phase 56 addendum 2 - full decomposition: the frontier's purge gain was a hidden per-scenario appetite dial; head reverted to legacy default
+
+Anchor decomposition on the purge (all paired, 48 scenarios):
+| arm | values (count source) | order | SCORE | vs old champ |
+|---|---|---|---|---|
+| unanchored frontier | logbag_rank | frontier | 1630.3 | -341.7 (publicly +215.6) |
+| scale-anchored | lbr shape, donor sums | frontier | 1925.1 | -46.9 +/- 42.6 (33W/14L, n.s.) |
+| value-anchored | plain logbag | frontier | 1934.8 | -37.2 +/- 44.1 (n.s.) |
+
+The smoking gun: mean swaps are FLAT (9.21 vs 9.23) but per-scenario counts
+swing |d| = 2.56 on average (23/48 scenarios differ by >= 2), and
+corr(count delta, cost gain) = 0.759. The 15 scenarios where logbag_rank's
+sum(p42) swelled the budget carry +956 gain each (~= the entire -342); the
+25 where it shrank carry -2. MECHANISM: in both-sides purge fits, future-side
+pairs from a deadly era raise the fitted base rate exactly in deadly
+scenarios - era-level hazard leaking through CALIBRATION. The label-window
+purge cannot block it; deployment's past-only fits cannot replicate it
+(measured collapse 0.78-0.80x at late origins); the public board refuted it
+(-63 swaps, +215.6). PURGE LIMITATION #3, and it retroactively weakens every
+"swaps flat at the mean" clearance in this campaign: the shipping bar is now
+COUNT-MATCHED (value-anchored) evaluation, where the frontier's true
+ordering value is -37..-47 +/- 43 - real-looking W/L (33/14) but sub-bar.
+
+ACTIONS: BATTERYSWAP_FRONTIER default OFF (head = legacy logbag champion,
+the measured-1661.45 behavior, plus all hardening); frontier dials retained
+for explicit A/B (ANCHOR=scale|value|off, W2). No model-side campaign
+candidate survives count-matched adjudication. Instruments going forward:
+deployment-mirror walk-forward (past-only fits) as the public predictor;
+building-GroupKFold with test-group-planned-as-fleet (user proposal,
+2026-08-22) as the held-out/private-fleet predictor.

batteryswap_example/submit.py

@@ -106,7 +106,13 @@ MODEL_KIND = 'logbag'
 # partner destroys the increment, so the public score is the arbiter).
 # Confirmed local optimum by 5 lanes (hyperparams, features, weights, neural
 # rankers, physics blends all subsumed or worse head-to-head).
-FRONTIER = os.environ.get('BATTERYSWAP_FRONTIER', '1') != '0'
+# DEFAULT OFF (phase 56 final decomposition): the frontier's purge gain was
+# a hidden per-scenario appetite dial - corr(count delta, cost gain) 0.76;
+# at matched counts the ordering is worth only -37..-47 +/- 43 (n.s.).
+# Era-hazard leaks through CALIBRATION in both-sides purge fits; deployment
+# (past-only) cannot replicate it and public refuted it (1877.08). The dials
+# below remain for explicit A/B only.
+FRONTIER = os.environ.get('BATTERYSWAP_FRONTIER', '0') != '0'
 # PUBLIC REFUTATION of the UNANCHORED frontier (698401a: 1877.08 vs 1661.45,
 # swaps 517 vs 580): under deployment's past-only fits, logbag_rank's
 # sum(p42) collapses vs plain logbag as the pool grows (measured ratio 1.02
@@ -116,7 +122,19 @@ FRONTIER = os.environ.get('BATTERYSWAP_FRONTIER', '1') != '0'
 # VALUE DONOR (count/pricing/timing = the measured-580-swap calibration,
 # byte-identical to the old champion's machinery) and the frontier supplies
 # ONLY the rank order. Ordering-only in deployment, not just in CV.
-FRONTIER_ANCHOR = os.environ.get('BATTERYSWAP_FRONTIER_ANCHOR', '1') != '0'
+# ANCHOR MODES (purge decomposition, phase 56 addendum 2): the frontier's
+# -341.7 CV gain is mostly the VALUE SHAPE, not the order - value-anchoring
+# to plain logbag keeps only the order and scores 1934.8 on the purge
+# (-37.2 +/- 44.1 vs old champion, n.s.). 'scale' keeps logbag_rank's value
+# SHAPE (the gain carrier) but rescales sum(p42) to the plain-logbag donor's
+# sum, so the swap count is the measured champion calibration by construction
+# while shape and order stay the frontier's. 'value' = donor values (order
+# only). 'off' = unanchored (publicly refuted 1877.08).
+FRONTIER_ANCHOR = os.environ.get('BATTERYSWAP_FRONTIER_ANCHOR', 'scale').lower()
+if FRONTIER_ANCHOR in ('1', 'true'):
+    FRONTIER_ANCHOR = 'value'
+if FRONTIER_ANCHOR in ('0', 'false', 'none'):
+    FRONTIER_ANCHOR = 'off'
 # Blend weight dial for public A/B: 0.30 = full frontier (purge 1630.3);
 # 0 = plain logbag_rank, no gbm (purge 1746.4 - the clean linear component,
 # fallback if the gbm increment turns out to be purge flattery on public);
@@ -142,8 +160,9 @@ class FrontierRUL:
         t['r_tau'] = t.groupby('origin')['tau_projected'].rank(pct=True)
         self.primary = LogBagFeats(['r_vlast', 'r_tau']).fit(t)
         self.partner = SklearnRUL('gbm').fit(t)
-        # value donor: the old champion's calibration owns the swap count
-        self.donor = SklearnRUL(MODEL_KIND).fit(train) if FRONTIER_ANCHOR else None
+        # donor: the old champion's calibration owns the swap count
+        self.donor = (SklearnRUL(MODEL_KIND).fit(train)
+                      if FRONTIER_ANCHOR != 'off' else None)
         return self
 
     def predict(self, rows):
@@ -157,12 +176,21 @@ class FrontierRUL:
         if FRONTIER_W2 > 0:
             p2 = self.partner.predict(r)
             rk = (1 - FRONTIER_W2) * rk + FRONTIER_W2 * p2['p42'].rank(pct=True)
-        base = self.donor.predict(rows) if self.donor is not None else pred
+        if FRONTIER_ANCHOR == 'value':
+            base = self.donor.predict(rows)      # donor values, frontier order
+        else:
+            base = pred                          # frontier values (and shape)
         vals = np.sort(base['p42'].values)[::-1]
         out = base.copy()
         out['p42'] = pd.Series(
             vals, index=rk.sort_values(ascending=False).index
         ).reindex(out.index)
+        if FRONTIER_ANCHOR == 'scale':
+            # calibration correction: frontier shape, donor COUNT - fixes the
+            # measured past-only sum(p42) collapse (0.78-0.80x at late origins)
+            dsum, psum = self.donor.predict(rows)['p42'].sum(), out['p42'].sum()
+            if psum > 0:
+                out['p42'] *= dsum / psum
         return out