Audit execution and add frozen research phases

README.md

@@ -36,9 +36,11 @@ missing candle, and writes a data snapshot plus manifest for every completed run
 - Walk-forward backtest uses training data only at each origin, checks the
   selected candle duration, skips gap-crossing targets, reports failures and
   coverage, and includes a last-close persistence baseline.
-- Backtests use the next candle's open as the entry, keep costs explicit, and
-  report separate 65% tuning and 35% final-check phases. The optional,
-  default-off random-signal control is available for leakage sanity checks.
+- Backtests use the next candle's open as the entry, keep costs explicit, prevent
+  overlapping positions, and expose separate chronological **60% TUNE / 20%
+  VALIDATE / 20% VAULT** phases. Only the requested phase is evaluated. The
+  optional, default-off random control runs 200 independent seeded executions
+  with matched trade counts and long/short proportions. Inversion is separate.
 - Live results are blocked when the latest closed candle is stale. The MAYTHOS
   heuristic and forecast metrics are shown separately.
 - Actionable live signals include causal Entry, Stop Loss, Take Profit and
@@ -60,6 +62,18 @@ missing candle, and writes a data snapshot plus manifest for every completed run
   profitable trade recommendations.
 - Every result can be downloaded as a ZIP containing candles, results, config,
   package versions, SHA-256 snapshot hash, and evaluation policy.
+- Win/loss and win-rate metrics use **net trade profit**, including profitable
+  time-stop exits and losing targets after costs. SL/TP/timeout outcome labels
+  remain separate from net-profit classification.
+- The **Frozen research** JSON API supports explicit historical endpoints,
+  immutable snapshot hashes, TUNE-only causal meta fitting, fixed per-trade
+  cash risk in R, doubled-cost stress and downloaded naive/null comparator rows.
+  Its stored forecast endpoint files are genuine historical Space inference
+  caches with provenance and hashes; the real pretrained model still loads.
+  They contain no trade outcomes or hardcoded performance values. A requested
+  snapshot hash mismatch stops evaluation.
+- Frozen research experiments have **not** demonstrated a winning configuration
+  under the admission gates. No profitability claim or winning default is made.
 
 ## Windows setup
 

app.py

@@ -139,7 +139,7 @@ def run_backtest_ui(*args):
             'Targets crossing gaps/market closures are explicitly skipped. Flat prices are classified as FLAT.\n\n'
             '**Accuracy denominator:** successful forecast evaluations; failures and coverage are shown separately. '
             '`persistence_mae` is the last-close naive baseline on the same valid targets. '
-            '**Trade execution is selective:** forecast targets must clear volatility noise, minimum 1.20R reward/risk, and 1.25× the configured round-trip costs; otherwise the forecast remains visible but no trade is opened. MAYTHOS uses only strong, unblocked signals. Candle-path resolution checks SL before TP when both touch; costs are explicit spread/fee/slippage assumptions, not provider-measured quotes.**\n\n'
+            f'**Trade execution is selective:** forecast targets must clear {config.minimum_edge_atr:g} ATR of volatility noise, minimum {config.minimum_risk_reward:g}R reward/risk, and 1.25× the configured round-trip costs; otherwise the forecast remains visible but no trade is opened. Forecast stops use {config.atr_stop_multiplier:g} ATR. MAYTHOS uses only strong, unblocked signals. Candle-path resolution checks SL before TP when both touch; costs are explicit spread/fee/slippage assumptions, not provider-measured quotes.**\n\n'
             + metadata_text(result['metadata'])
         )
         artifact = export_run(result, 'backtest')

risk_management.py

@@ -220,7 +220,7 @@ def build_trade_levels(
     except (TypeError, ValueError):
         forecast_value = None
     if any(isinstance(value, (bool, np.bool_)) for value in (
-            minimum_edge_atr, minimum_risk_reward, minimum_edge_bps)):
+            minimum_edge_atr, minimum_risk_reward, minimum_edge_bps, atr_stop_multiplier)):
         raise ValueError('minimum_edge_atr/minimum_edge_bps must be numeric and minimum_risk_reward must be numeric.')
     try:
         minimum_edge_atr = float(minimum_edge_atr)

tests/test_audit_regressions.py

@@ -68,9 +68,9 @@ def test_win_loss_counts_reconcile_with_net_profit():
     frame = pd.DataFrame({'trade_taken': [True]*3, 'net_r': [-0.1, 0.3, -1.0],
                           'outcome': ['TP', 'TIMEOUT', 'SL']})
     summary = summarize_trades(frame)
-    assert summary['wins'] == 1
+    assert summary['wins'] == 2
     assert summary['losses'] == 1
-    assert summary['win_rate_pct'] == pytest.approx(33.33)
+    assert summary['win_rate_pct'] == pytest.approx(66.67)
 
 
 def test_invalid_intermediate_forecast_is_not_scored():