responsive and representative demand for COI erosion

engine/lib/demand.py

@@ -17,18 +17,32 @@ def generate_demand_for_actor(
     params: tuple,
     noise_std: float = 1.0,
     distribution_method=np.random.normal,
+    normalize: bool = False,
 ) -> np.ndarray:
     """d(p;0) = max(0, valuation - price) + epsi for single actor type
     params: (mean, std) for valuation distribution D_H or D_A"""
     val = distribution_method(*params, size=len(prices))
     noise = distribution_method(0, noise_std, len(prices))
     demand = np.maximum(0, val - prices + noise)
+    if not normalize:
+        return demand
     total = np.sum(demand)
     return demand / total * 100 if total > 0 else demand
 
 
-def estimate_demand(trajectories, action_weights=None):
-    return estimate_weighted_demand(trajectories, action_weights)
+def estimate_demand(
+    trajectories,
+    action_weights=None,
+    *,
+    normalize: bool = False,
+    per_session: bool = True,
+):
+    return estimate_weighted_demand(
+        trajectories,
+        action_weights,
+        normalize=normalize,
+        per_session=per_session,
+    )
 
 
 def _parse_event_state(state: str):
@@ -50,7 +64,13 @@ def _weight_for_action(action: str, action_weights: dict) -> float:
     return CATEGORY_WEIGHTS["nav"]
 
 
-def estimate_weighted_demand(trajectories, action_weights=None):
+def estimate_weighted_demand(
+    trajectories,
+    action_weights=None,
+    *,
+    normalize: bool = False,
+    per_session: bool = True,
+):
     action_weights = (
         DEFAULT_ACTION_WEIGHTS if action_weights is None else action_weights
     )
@@ -64,12 +84,20 @@ def estimate_weighted_demand(trajectories, action_weights=None):
             if w <= 0:
                 continue
             scores[product_id] = scores.get(product_id, 0.0) + w
-    total = sum(scores.values())
-    return (
-        {pid: (score / total) * 100 for pid, score in scores.items()}
-        if total > 0
-        else {}
-    )
+    if not scores:
+        return {}
+
+    if per_session and len(trajectories) > 0:
+        inv_n = 1.0 / float(len(trajectories))
+        scores = {pid: score * inv_n for pid, score in scores.items()}
+
+    if not normalize:
+        return scores
+
+    total = float(sum(scores.values()))
+    if total <= 0:
+        return {}
+    return {pid: (score / total) * 100.0 for pid, score in scores.items()}
 
 
 # Example usage

engine/lib/wrappers.py

@@ -32,17 +32,23 @@ class EconomicMetricsWrapper(gym.Wrapper):
         obs, reward, terminated, truncated, info = self.env.step(action)
 
         # extract from unwrapped env
-        prices = self.env.unwrapped._prices
+        quoted_prices = np.asarray(self.env.unwrapped._prices, dtype=float)
+        effective_prices = np.asarray(
+            info.get("effective_prices", quoted_prices), dtype=float
+        )
+        if effective_prices.shape != quoted_prices.shape:
+            effective_prices = quoted_prices
         demand_dict = self.env.unwrapped._demand
-        demand = np.array([demand_dict.get(i, 0.0) for i in range(len(prices))])
+        demand = np.array([demand_dict.get(i, 0.0) for i in range(len(quoted_prices))])
 
         # core calculations
-        revenue = float(np.sum(prices * demand))
-        avg_price = float(np.mean(prices))
+        revenue = float(info.get("revenue", np.sum(effective_prices * demand)))
+        quoted_revenue = float(np.sum(quoted_prices * demand))
+        avg_price = float(np.mean(effective_prices))
         margin = (avg_price - self.p_min) / max(avg_price, 1e-6)
         coi_level = avg_price - self.p_min  # E[P] - p_min per thesis Def 1
 
-        self._price_history.append(prices.copy())
+        self._price_history.append(effective_prices.copy())
         self._revenue_history.append(revenue)
 
         # regret vs baseline (golden path)
@@ -53,6 +59,7 @@ class EconomicMetricsWrapper(gym.Wrapper):
         # inject structured metrics into info
         info["economics"] = {
             "revenue": revenue,
+            "quoted_revenue": quoted_revenue,
             "margin": margin,
             "coi_level": coi_level,
             "regret": regret,
@@ -71,10 +78,13 @@ class EconomicMetricsWrapper(gym.Wrapper):
             "agent_prob",
             "alpha_adv",
             "alpha_nominal",
+            "erosion_share",
+            "effective_price_mean",
         ):
             if key in info:
                 info["economics"][key] = info[key]
-        info["prices"] = prices.copy()
+        info["prices"] = quoted_prices.copy()
+        info["effective_prices"] = effective_prices.copy()
         info["demand"] = demand.copy()
 
         return obs, reward, terminated, truncated, info