Add updated files

game_data.py

@@ -1,4 +1,4 @@
-# Model Baselines (per 1000 tokens) based on scaling laws [cite: 26, 27]
+# Model Baselines (per 1000 tokens) based on scaling laws
 MODELS = {
     "Mistral Large 3": {"energy_wh": 1.98, "water_ml": 3.56, "co2_g": 0.95},
     "Mistral Medium 3": {"energy_wh": 6.76, "water_ml": 12.17, "co2_g": 3.24},
@@ -10,6 +10,7 @@ MODELS = {
     "GPT-5.4": {"energy_wh": 169.05, "water_ml": 304.29, "co2_g": 81.14}
 }
 
+# Category Token Multipliers
 CATEGORY_MULTIPLIERS = {
     "simple factual question": 0.05,        
     "mathematical calculation": 0.10,       
@@ -18,14 +19,26 @@ CATEGORY_MULTIPLIERS = {
 }
 
 def calculate_impact(category, confidence_score, query_text, model_name):
+    # Get base costs for the specific model
     model_cost = MODELS.get(model_name, MODELS["Mistral Large 3"])
     cat_mult = CATEGORY_MULTIPLIERS.get(category, 0.05)
     
-    # Logic for buffers and length factors [cite: 25, 29]
-    conf_mult = 1.0 if confidence_score > 0.8 else (1.2 if confidence_score > 0.5 else 1.5)
+    # 1. Confidence Multiplier
+    if confidence_score > 0.8:
+        conf_mult = 1.0
+    elif 0.5 <= confidence_score <= 0.8:
+        conf_mult = 1.2
+    else:
+        conf_mult = 1.5
+        
+    # 2. Dynamic Word Count Logic (Continuous Scaling)
     word_count = len(query_text.split())
-    len_factor = 1.0 if word_count < 20 else (1.3 if word_count <= 50 else 1.5)
     
+    # Baseline is 15 words (1.0x). Each word adjusts the impact by 1.5% (0.015)
+    # Minimum factor is floored at 0.5x to ensure small queries still cost resources
+    len_factor = max(0.5, 1.0 + ((word_count - 15) * 0.015))
+    
+    # 3. Final Calculation Formula
     water_ml = model_cost["water_ml"] * cat_mult * conf_mult * len_factor
     energy_wh = model_cost["energy_wh"] * cat_mult * conf_mult * len_factor
     co2_g = model_cost["co2_g"] * cat_mult * conf_mult * len_factor