adding redistribution bonus

app.py

@@ -8,7 +8,7 @@ from pathlib import Path
 import pandas as pd
 import pytz
 import streamlit as st
-from datasets import Dataset, load_dataset, Features, Value
+from datasets import Dataset, load_dataset
 from huggingface_hub import CommitScheduler
 
 # File paths as constants
@@ -303,7 +303,7 @@ def display_predictions():
 def display_leaderboard():
     if st.button("Show Leaderboard"):
         try:
-            # # Load the 'leaders' configuration
+            # Load the 'leaders' configuration
             dataset = load_dataset("Jay-Rajput/DIS_IPL_Leads", split='train')
 
             users_data = []
@@ -311,11 +311,11 @@ def display_leaderboard():
                 for user, points_dict in dataset[0].items():
                     points = points_dict.get("points", 0)
                     last_5_results = " ".join(points_dict.get("last_5_results", ["⚪"] * 5))  # Default: 5 white circles
-                    # bonus = points_dict.get("redistributed_bonus", 0)
+                    bonus = points_dict.get("redistributed_bonus", 0)
                     users_data.append({
                         'User': user,
                         'Points': points,
-                        # 'Redistribution Bonus': bonus,
+                        'Redistribution Bonus': bonus,
                         'Last 5 Bids': last_5_results
                     })
             else:
@@ -414,7 +414,6 @@ ADMIN_PASSPHRASE = "admin123"
 
 
 def fetch_latest_predictions(match_id):
-    # Load the dataset. Adjust "split" to "train" or appropriate if "predictions" is a configuration.
     dataset = load_dataset("Jay-Rajput/DIS_IPL_Preds", split="train")
     # Convert the dataset to a pandas DataFrame
     df = pd.DataFrame(dataset)
@@ -492,12 +491,10 @@ def redistribute_lost_points(match_id):
 
 
 def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match, outcome_only=False):
-
-    # Load existing match outcomes
     outcomes = load_dataset("Jay-Rajput/DIS_IPL_Outcomes", split="train")  
     outcomes_df = pd.DataFrame(outcomes)
 
-    # Directly update or add the match outcome
+    # Update or add match outcome
     outcome_exists = False
     for idx, outcome in outcomes_df.iterrows():
         if outcome['match_id'] == match_id:
@@ -510,65 +507,191 @@ def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match, ou
         outcomes_df = pd.concat([outcomes_df, pd.DataFrame([new_outcome])], ignore_index=True)
         outcomes = Dataset.from_pandas(outcomes_df)
 
-    if not outcome_only:  # Update user scores only if outcome_only is False
-        # Load predictions only if necessary
+    if not outcome_only:
         predictions = fetch_latest_predictions(match_id)
-
-        # Load users' data only if necessary
         users = load_dataset("Jay-Rajput/DIS_IPL_Leads", split="train")
         users_df = pd.DataFrame(users)
 
-        # Update user points based on prediction accuracy
-        users_with_predictions = set(predictions['user_name'])
+        # Capture previous leaderboard (top 5 users and their points)
+        prev_scores = [(user, users_df[user][0]['points']) for user in users_df.columns]
+        prev_scores.sort(key=lambda x: x[1], reverse=True)
+        prev_top_5 = prev_scores[:5]
+        top5_usernames = [user for user, _ in prev_top_5]
+
+        lost_points_by_top5 = 0
+        user_outcomes = {}
+
+        # Step 1: Apply current match outcomes
         for user_name in users_df.columns:
-            user_points = users_df[user_name][0]['points']
-            if user_name in users_with_predictions:
+            user_data = users_df[user_name][0]
+            user_points = user_data['points']
+            user_initial_points = user_points
+
+            if user_name in set(predictions['user_name']):
                 prediction = predictions[predictions['user_name'] == user_name].iloc[0]
                 predicted_winner = prediction['predicted_winner']
                 predicted_motm = prediction['predicted_motm']
                 bid_points = prediction['bid_points']
 
-                # Update points based on prediction accuracy
                 if predicted_winner == winning_team:
                     user_points += 2000 + bid_points
-                    result_indicator = "🟢"  # Correct Prediction
+                    result_indicator = "🟢"
                     if predicted_motm == man_of_the_match:
                         user_points += 500
                 else:
                     user_points -= 200 + bid_points
-                    result_indicator = "🔴"  # Wrong Prediction
+                    result_indicator = "🔴"
+                    if user_name in top5_usernames:
+                        lost_points_by_top5 += (200 + bid_points)
             else:
-                # Deduct 200 points for not submitting a prediction
                 user_points -= 200
-                result_indicator = "⚪"  # No Prediction
-                
-            # Ensure user_points is never negative
+                result_indicator = "⚪"
+                if user_name in top5_usernames:
+                    lost_points_by_top5 += 200
+
             user_points = max(user_points, 0)
+            user_outcomes[user_name] = {
+                "updated_points": user_points,
+                "result_indicator": result_indicator,
+                "initial_points": user_initial_points
+            }
+
+        # Step 2: Build new leaderboard after applying outcome
+        new_leaderboard = [(u, d["updated_points"]) for u, d in user_outcomes.items()]
+        new_leaderboard.sort(key=lambda x: x[1], reverse=True)
+        new_ranks = {user: rank for rank, (user, _) in enumerate(new_leaderboard)}
+
+        # Step 3: Try redistributing lost points to remaining users
+        remaining_users = [u for u in users_df.columns if u not in top5_usernames]
+        redistribution_possible = True
+        redistributed_outcomes = {}
+
+        if remaining_users and lost_points_by_top5 > 0:
+            share = lost_points_by_top5 // len(remaining_users)
+
+            # Simulate redistribution
+            for user in user_outcomes:
+                redistributed_outcomes[user] = user_outcomes[user]["updated_points"]
+                if user in remaining_users:
+                    redistributed_outcomes[user] += share
+
+            # Check if new ranks remain same
+            redistributed_leaderboard = [(u, p) for u, p in redistributed_outcomes.items()]
+            redistributed_leaderboard.sort(key=lambda x: x[1], reverse=True)
+            redistributed_ranks = {user: rank for rank, (user, _) in enumerate(redistributed_leaderboard)}
+
+            for user in user_outcomes:
+                if new_ranks[user] != redistributed_ranks[user]:
+                    redistribution_possible = False
+                    break
+
+        # Step 4: Apply the appropriate update
+        for user in users_df.columns:
+            if redistribution_possible and user in redistributed_outcomes:
+                final_points = redistributed_outcomes[user]
+                bonus = final_points - user_outcomes[user]["updated_points"]
+            else:
+                final_points = user_outcomes[user]["updated_points"]
+                bonus = 0  # No redistribution received
 
-            # Update user's points in the DataFrame
-            users_df[user_name][0]['points'] = user_points
-            users[user_name][0]['points'] = user_points
-            
-            # Maintain last 5 prediction results
-            if "last_5_results" not in users_df[user_name][0]:
-                users_df[user_name][0]["last_5_results"] = []
-            users_df[user_name][0]["last_5_results"].insert(0, result_indicator)  # Insert at beginning
-            users_df[user_name][0]["last_5_results"] = users_df[user_name][0]["last_5_results"][:5]  # Keep only last 5
-            
-            if "last_5_results" not in users[user_name][0]:
-                users[user_name][0]["last_5_results"] = []
-            users[user_name][0]["last_5_results"].insert(0, result_indicator)  # Insert at beginning
-            users[user_name][0]["last_5_results"] = users[user_name][0]["last_5_results"][:5]  # Keep only last 5
+            users_df[user][0]['points'] = final_points
+            users_df[user][0]['redistributed_bonus'] = bonus  # Track redistributed bonus
+
+            # Maintain last 5 results
+            result = user_outcomes[user]["result_indicator"]
+            if "last_5_results" not in users_df[user][0]:
+                users_df[user][0]["last_5_results"] = []
+            users_df[user][0]["last_5_results"].insert(0, result)
+            users_df[user][0]["last_5_results"] = users_df[user][0]["last_5_results"][:5]
 
+        # Save updated leaderboard
         users.to_json(USERS_JSON)
         updated_dataset = Dataset.from_pandas(users_df)
         updated_dataset.push_to_hub("Jay-Rajput/DIS_IPL_Leads", split="train")
-        # redistribute_lost_points(match_id)
 
+    # Save match outcome
     outcomes.to_json(OUTCOMES)
     outcomes.push_to_hub("Jay-Rajput/DIS_IPL_Outcomes", split="train")
 
 
+# def update_leaderboard_and_outcomes(match_id, winning_team, man_of_the_match, outcome_only=False):
+
+#     # Load existing match outcomes
+#     outcomes = load_dataset("Jay-Rajput/DIS_IPL_Outcomes", split="train")  
+#     outcomes_df = pd.DataFrame(outcomes)
+
+#     # Directly update or add the match outcome
+#     outcome_exists = False
+#     for idx, outcome in outcomes_df.iterrows():
+#         if outcome['match_id'] == match_id:
+#             outcomes_df.at[idx, 'winning_team'] = winning_team
+#             outcomes_df.at[idx, 'man_of_the_match'] = man_of_the_match
+#             outcome_exists = True
+#             break
+#     if not outcome_exists:
+#         new_outcome = {"match_id": match_id, "winning_team": winning_team, "man_of_the_match": man_of_the_match}
+#         outcomes_df = pd.concat([outcomes_df, pd.DataFrame([new_outcome])], ignore_index=True)
+#         outcomes = Dataset.from_pandas(outcomes_df)
+
+#     if not outcome_only:  # Update user scores only if outcome_only is False
+#         # Load predictions only if necessary
+#         predictions = fetch_latest_predictions(match_id)
+
+#         # Load users' data only if necessary
+#         users = load_dataset("Jay-Rajput/DIS_IPL_Leads", split="train")
+#         users_df = pd.DataFrame(users)
+
+#         # Update user points based on prediction accuracy
+#         users_with_predictions = set(predictions['user_name'])
+#         for user_name in users_df.columns:
+#             user_points = users_df[user_name][0]['points']
+#             if user_name in users_with_predictions:
+#                 prediction = predictions[predictions['user_name'] == user_name].iloc[0]
+#                 predicted_winner = prediction['predicted_winner']
+#                 predicted_motm = prediction['predicted_motm']
+#                 bid_points = prediction['bid_points']
+
+#                 # Update points based on prediction accuracy
+#                 if predicted_winner == winning_team:
+#                     user_points += 2000 + bid_points
+#                     result_indicator = "🟢"  # Correct Prediction
+#                     if predicted_motm == man_of_the_match:
+#                         user_points += 500
+#                 else:
+#                     user_points -= 200 + bid_points
+#                     result_indicator = "🔴"  # Wrong Prediction
+#             else:
+#                 # Deduct 200 points for not submitting a prediction
+#                 user_points -= 200
+#                 result_indicator = "⚪"  # No Prediction
+                
+#             # Ensure user_points is never negative
+#             user_points = max(user_points, 0)
+
+#             # Update user's points in the DataFrame
+#             users_df[user_name][0]['points'] = user_points
+#             users[user_name][0]['points'] = user_points
+            
+#             # Maintain last 5 prediction results
+#             if "last_5_results" not in users_df[user_name][0]:
+#                 users_df[user_name][0]["last_5_results"] = []
+#             users_df[user_name][0]["last_5_results"].insert(0, result_indicator)  # Insert at beginning
+#             users_df[user_name][0]["last_5_results"] = users_df[user_name][0]["last_5_results"][:5]  # Keep only last 5
+            
+#             if "last_5_results" not in users[user_name][0]:
+#                 users[user_name][0]["last_5_results"] = []
+#             users[user_name][0]["last_5_results"].insert(0, result_indicator)  # Insert at beginning
+#             users[user_name][0]["last_5_results"] = users[user_name][0]["last_5_results"][:5]  # Keep only last 5
+
+#         users.to_json(USERS_JSON)
+#         updated_dataset = Dataset.from_pandas(users_df)
+#         updated_dataset.push_to_hub("Jay-Rajput/DIS_IPL_Leads", split="train")
+#         # redistribute_lost_points(match_id)
+
+#     outcomes.to_json(OUTCOMES)
+#     outcomes.push_to_hub("Jay-Rajput/DIS_IPL_Outcomes", split="train")
+
+
 # Function to fetch matches for a given date
 def fetch_matches_by_date(matches, selected_date):
     return [match for match in matches if datetime.strptime(match['date'], '%Y-%m-%d').date() == selected_date]