fix: 4 critical optimizer bugs

- NaN composite score: dropna().mean() on empty series returns NaN,
  nan is truthy so (nan or 0)=nan. Fixed: use None when no MFE data
  so scorer uses 0.5 fallback instead of propagating NaN.

- Analysis stopping: gate WFV was importing from main.execute_run
  (old CLI code that uses broken runner). Fixed: gate.run_walk_forward
  now accepts an executor callable from optimizer_loop.

- Back to Dashboard opened blank tab: Reports button used
  window.open(_blank). Fixed: same-tab navigation + history.back().

- NaN in reports card: downstream of composite_score NaN above.

optimizer_loop.py

@@ -268,11 +268,24 @@ class OptimizerLoop:
                 no_improve_count += 1
                 continue
 
-            # Walk-forward
+            # Walk-forward — pass an executor so gate never imports 'main'
             self._emit("log", {"level": "info", "msg": "Running walk-forward validation..."})
+            _store, _builder, _runner, _parser, _log_rdr, _analyzers, _scorer = (
+                store, builder, runner, parser, log_rdr, analyzers, scorer
+            )
+            def _wfv_executor(params, start, end, fold_id):
+                m, _ = self._execute_run(
+                    run_id=fold_id, params=params,
+                    period_start=start, period_end=end,
+                    phase="wfv", hypothesis_id=None,
+                    store=_store, builder=_builder, runner=_runner,
+                    parser=_parser, log_rdr=_log_rdr,
+                    analyzers=_analyzers, scorer=_scorer,
+                )
+                return m
             wfv = gate.run_walk_forward(
-                iteration_best_params, cfg, store, builder, runner,
-                parser, log_rdr, analyzers, scorer,
+                params=iteration_best_params,
+                executor=_wfv_executor,
             )
             self._emit("log", {
                 "level": "success" if wfv.passed else "warn",
@@ -406,7 +419,10 @@ class OptimizerLoop:
                 reversals = trades_df[trades_df["result_class"] == "reversal"]
                 metrics.reversal_rate = len(reversals) / max(1, len(losers))
             if "mfe_capture_ratio" in trades_df.columns:
-                metrics.avg_mfe_capture = float(trades_df["mfe_capture_ratio"].dropna().mean() or 0)
+                # dropna() first — if no MFE data, series is all-NaN, mean()=NaN
+                # Use None (not NaN) so scorer uses its safe default of 0.5
+                cap_series = trades_df["mfe_capture_ratio"].dropna()
+                metrics.avg_mfe_capture = float(cap_series.mean()) if not cap_series.empty else None
 
         metrics.composite_score = scorer.score(metrics)
         store.save_metrics(metrics)

ui/templates/index.html

@@ -39,7 +39,7 @@
       </div>
     </div>
     <div class="header-right">
-      <button class="btn btn-secondary" id="btn-reports" onclick="window.open('/reports','_blank')">
+      <button class="btn btn-secondary" id="btn-reports" onclick="window.location.href='/reports'">
         📁 Reports
       </button>
       <button class="btn btn-pause hidden" id="btn-pause" onclick="pauseOptimizer()">

ui/templates/reports_index.html

@@ -38,7 +38,7 @@
       <h1>📁 Optimization Reports</h1>
       <div class="sub">{{ runs|length }} run(s) recorded · Click any card to open the full report</div>
     </div>
-    <a class="back-link" href="/">← Back to Dashboard</a>
+    <a class="back-link" href="/" onclick="if(document.referrer){{history.back();return false;}}">← Back to Dashboard</a>
   </header>
 
   {% if not runs %}

validation/gate.py

@@ -60,16 +60,9 @@ class ValidationGate:
 
     def run_walk_forward(
         self,
-        params:     dict[str, Any],
-        cfg:        dict,
-        store,       # DataStore
-        builder,     # IniBuilder
-        runner,      # MT5Runner
-        parser,      # ReportParser
-        log_rdr,     # TradeLogReader
-        analyzers,   # list[BaseAnalyzer]
-        scorer,      # CompositeScorer
-        n_folds:    int = 2,
+        params:   dict[str, Any],
+        executor,                 # callable(params, start_str, end_str, fold_id) -> Optional[RunMetrics]
+        n_folds:  int = 2,
     ) -> WFVResult:
         """
         Split training period into n_folds sub-periods.
@@ -94,21 +87,13 @@ class ValidationGate:
             if i == n_folds - 1:
                 fold_end = train_end   # last fold gets remainder
 
-            fold_id = f"wfv_fold{i+1}_{uuid.uuid4().hex[:6]}"
             logger.info(f"WFV fold {i+1}/{n_folds}: {fold_start.date()} → {fold_end.date()}")
 
-            # Import here to avoid circular
-            from main import execute_run
-            fm, _ = execute_run(
-                run_id=fold_id,
-                params=params,
-                period_start=fold_start.strftime("%Y.%m.%d"),
-                period_end=fold_end.strftime("%Y.%m.%d"),
-                phase="wfv",
-                hypothesis_id=None,
-                cfg=cfg, store=store, builder=builder, runner=runner,
-                parser=parser, log_rdr=log_rdr, analyzers=analyzers, scorer=scorer,
-            )
+            fold_id = f"wfv_fold{i+1}_{uuid.uuid4().hex[:6]}"
+            start_s = fold_start.strftime("%Y.%m.%d")
+            end_s   = fold_end.strftime("%Y.%m.%d")
+
+            fm = executor(params, start_s, end_s, fold_id)
             if fm:
                 fold_metrics.append(fm)