global_func/predict_dupes.py

import streamlit as st
import numpy as np
import pandas as pd
import time
from fuzzywuzzy import process
import math
from difflib import SequenceMatcher

def calculate_weighted_ownership(row_ownerships):
    """
    Calculate weighted ownership based on the formula:
    (AVERAGE of (each value's average with overall average)) * count - (max - min)
    
    Args:
        row_ownerships: Series containing ownership values in percentage form (e.g., 24.2213 for 24.2213%)
        
    Returns:
        float: Calculated weighted ownership value
    """
    # Drop NaN values and convert percentages to decimals
    row_ownerships = row_ownerships.dropna() / 100
    
    # Get the mean of all ownership values
    row_mean = row_ownerships.mean()
    
    # Calculate average of each value with the overall mean
    value_means = [(val + row_mean) / 2 for val in row_ownerships]
    
    # Take average of all those means
    avg_of_means = sum(value_means) / len(row_ownerships)
    
    # Multiply by count of values
    weighted = avg_of_means * (len(row_ownerships) * 1)
    
    # Subtract (max - min)
    weighted = weighted - (row_ownerships.max() - row_ownerships.min())
    
    # Convert back to percentage form to match input format
    return weighted * 10000

def calculate_player_similarity_score(portfolio, player_columns):
    """
    Calculate a similarity score that measures how different each row is from all other rows
    based on actual player selection. Optimized for speed using vectorized operations.
    Higher scores indicate more unique/different lineups.
    
    Args:
        portfolio: DataFrame containing the portfolio data
        player_columns: List of column names containing player names
        
    Returns:
        Series: Similarity scores for each row
    """
    # Extract player data
    player_data = portfolio[player_columns].fillna('')
    
    # Get all unique players and create a mapping to numeric IDs
    all_players = set()
    for col in player_columns:
        unique_vals = player_data[col].unique()
        for val in unique_vals:
            if isinstance(val, str) and val.strip() != '':
                all_players.add(val)
    
    # Create player ID mapping
    player_to_id = {player: idx for idx, player in enumerate(sorted(all_players))}
    
    # Convert each row to a binary vector (1 if player is present, 0 if not)
    n_players = len(all_players)
    n_rows = len(portfolio)
    binary_matrix = np.zeros((n_rows, n_players), dtype=np.int8)
    
    for i, (_, row) in enumerate(player_data.iterrows()):
        for val in row.values:
            if isinstance(val, str) and str(val).strip() != '' and str(val) in player_to_id:
                binary_matrix[i, player_to_id[str(val)]] = 1
    
    # Vectorized Jaccard distance calculation
    # Use matrix operations to compute all pairwise distances at once
    similarity_scores = np.zeros(n_rows)
    
    # Compute intersection and union matrices
    # intersection[i,j] = number of players in common between row i and row j
    # union[i,j] = total number of unique players between row i and row j
    intersection_matrix = np.dot(binary_matrix, binary_matrix.T)
    
    # For union, we need: |A ∪ B| = |A| + |B| - |A ∩ B|
    row_sums = np.sum(binary_matrix, axis=1)
    union_matrix = row_sums[:, np.newaxis] + row_sums - intersection_matrix
    
    # Calculate Jaccard distance: 1 - (intersection / union)
    # Avoid division by zero
    with np.errstate(divide='ignore', invalid='ignore'):
        jaccard_similarity = np.divide(intersection_matrix, union_matrix, 
                                     out=np.zeros_like(intersection_matrix, dtype=float), 
                                     where=union_matrix != 0)
    
    # Convert similarity to distance and calculate average distance for each row
    jaccard_distance = 1 - jaccard_similarity
    
    # For each row, calculate average distance to all other rows
    # Exclude self-comparison (diagonal elements)
    np.fill_diagonal(jaccard_distance, 0)
    row_counts = n_rows - 1  # Exclude self
    similarity_scores = np.sum(jaccard_distance, axis=1) / row_counts
    
    # Normalize to 0-1 scale where 1 = most unique/different
    if similarity_scores.max() > similarity_scores.min():
        similarity_scores = (similarity_scores - similarity_scores.min()) / (similarity_scores.max() - similarity_scores.min())
    
    return similarity_scores

def predict_dupes(portfolio, maps_dict, site_var, type_var, Contest_Size, strength_var, sport_var):
    if strength_var == 'Weak':
        dupes_multiplier = .75
        percentile_multiplier = .90
    elif strength_var == 'Average':
        dupes_multiplier = 1.00
        percentile_multiplier = 1.00
    elif strength_var == 'Sharp':
        dupes_multiplier = 1.25
        percentile_multiplier = 1.10
    max_ownership = max(maps_dict['own_map'].values()) / 100
    average_ownership = np.mean(list(maps_dict['own_map'].values())) / 100
    if site_var == 'Fanduel':
        if type_var == 'Showdown':
            dup_count_columns = ['CPT_Own_percent_rank', 'FLEX1_Own_percent_rank', 'FLEX2_Own_percent_rank', 'FLEX3_Own_percent_rank', 'FLEX4_Own_percent_rank']
            own_columns = ['CPT_Own', 'FLEX1_Own', 'FLEX2_Own', 'FLEX3_Own', 'FLEX4_Own']
            calc_columns = ['own_product', 'own_average', 'own_sum', 'avg_own_rank', 'dupes_calc', 'low_own_count', 'own_ratio', 'Ref_Proj', 'Max_Proj', 'Min_Proj', 'Avg_Ref', 'own_ratio']
            # Get the original player columns (first 5 columns excluding salary, median, Own)
            player_columns = [col for col in portfolio.columns[:5] if col not in ['salary', 'median', 'Own']]
            
            flex_ownerships = pd.concat([
                portfolio.iloc[:,1].map(maps_dict['own_map']),
                portfolio.iloc[:,2].map(maps_dict['own_map']),
                portfolio.iloc[:,3].map(maps_dict['own_map']),
                portfolio.iloc[:,4].map(maps_dict['own_map'])
            ])
            flex_rank = flex_ownerships.rank(pct=True)
            
            # Assign ranks back to individual columns using the same rank scale
            portfolio['CPT_Own_percent_rank'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']).rank(pct=True)
            portfolio['FLEX1_Own_percent_rank'] = portfolio.iloc[:,1].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX2_Own_percent_rank'] = portfolio.iloc[:,2].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX3_Own_percent_rank'] = portfolio.iloc[:,3].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX4_Own_percent_rank'] = portfolio.iloc[:,4].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])

            portfolio['CPT_Own'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']) / 100
            portfolio['FLEX1_Own'] = portfolio.iloc[:,1].map(maps_dict['own_map']) / 100
            portfolio['FLEX2_Own'] = portfolio.iloc[:,2].map(maps_dict['own_map']) / 100
            portfolio['FLEX3_Own'] = portfolio.iloc[:,3].map(maps_dict['own_map']) / 100
            portfolio['FLEX4_Own'] = portfolio.iloc[:,4].map(maps_dict['own_map']) / 100
            
            portfolio['own_product'] = (portfolio[own_columns].product(axis=1))
            portfolio['own_average'] = (portfolio['Own'].max() * .33) / 100
            portfolio['own_sum'] = portfolio[own_columns].sum(axis=1)
            portfolio['avg_own_rank'] = portfolio[dup_count_columns].mean(axis=1)
            
            # Calculate dupes formula
            portfolio['dupes_calc'] = (portfolio['own_product'] * portfolio['avg_own_rank']) * Contest_Size + ((portfolio['salary'] - (60000 - portfolio['Own'])) / 100) - ((60000 - portfolio['salary']) / 100)
            portfolio['dupes_calc'] = portfolio['dupes_calc'] * dupes_multiplier
            
            # Round and handle negative values
            portfolio['Dupes'] = np.where(
                np.round(portfolio['dupes_calc'], 0) <= 0,
                0, 
                np.round(portfolio['dupes_calc'], 0) - 1
            )
        elif type_var == 'Classic':
            num_players = len([col for col in portfolio.columns if col not in ['salary', 'median', 'Own']])
            dup_count_columns = [f'player_{i}_percent_rank' for i in range(1, num_players + 1)]
            own_columns = [f'player_{i}_own' for i in range(1, num_players + 1)]
            calc_columns = ['own_product', 'own_average', 'own_sum', 'avg_own_rank', 'dupes_calc', 'low_own_count', 'own_ratio', 'Ref_Proj', 'Max_Proj', 'Min_Proj', 'Avg_Ref', 'own_ratio']
            # Get the original player columns (first num_players columns excluding salary, median, Own)
            player_columns = [col for col in portfolio.columns[:num_players] if col not in ['salary', 'median', 'Own']]
            
            for i in range(1, num_players + 1):
                portfolio[f'player_{i}_percent_rank'] = portfolio.iloc[:,i-1].map(maps_dict['own_percent_rank'])
                portfolio[f'player_{i}_own'] = portfolio.iloc[:,i-1].map(maps_dict['own_map']) / 100
            
            portfolio['own_product'] = (portfolio[own_columns].product(axis=1))
            portfolio['own_average'] = (portfolio['Own'].max() * .33) / 100
            portfolio['own_sum'] = portfolio[own_columns].sum(axis=1)
            portfolio['avg_own_rank'] = portfolio[dup_count_columns].mean(axis=1)
            
            portfolio['dupes_calc'] = (portfolio['own_product'] * portfolio['avg_own_rank']) * Contest_Size + ((portfolio['salary'] - (60000 - portfolio['Own'])) / 100) - ((60000 - portfolio['salary']) / 100)
            portfolio['dupes_calc'] = portfolio['dupes_calc'] * dupes_multiplier
            # Round and handle negative values
            portfolio['Dupes'] = np.where(
                np.round(portfolio['dupes_calc'], 0) <= 0,
                0, 
                np.round(portfolio['dupes_calc'], 0) - 1
            )

    elif site_var == 'Draftkings':
        if type_var == 'Showdown':
            dup_count_columns = ['CPT_Own_percent_rank', 'FLEX1_Own_percent_rank', 'FLEX2_Own_percent_rank', 'FLEX3_Own_percent_rank', 'FLEX4_Own_percent_rank', 'FLEX5_Own_percent_rank']
            own_columns = ['CPT_Own', 'FLEX1_Own', 'FLEX2_Own', 'FLEX3_Own', 'FLEX4_Own', 'FLEX5_Own']
            calc_columns = ['own_product', 'own_average', 'own_sum', 'avg_own_rank', 'dupes_calc', 'low_own_count', 'Ref_Proj', 'Max_Proj', 'Min_Proj', 'Avg_Ref', 'own_ratio']
            # Get the original player columns (first 6 columns excluding salary, median, Own)
            player_columns = [col for col in portfolio.columns[:6] if col not in ['salary', 'median', 'Own']]
            
            flex_ownerships = pd.concat([
                portfolio.iloc[:,1].map(maps_dict['own_map']),
                portfolio.iloc[:,2].map(maps_dict['own_map']),
                portfolio.iloc[:,3].map(maps_dict['own_map']),
                portfolio.iloc[:,4].map(maps_dict['own_map']),
                portfolio.iloc[:,5].map(maps_dict['own_map'])
            ])
            flex_rank = flex_ownerships.rank(pct=True)
            
            # Assign ranks back to individual columns using the same rank scale
            portfolio['CPT_Own_percent_rank'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']).rank(pct=True)
            portfolio['FLEX1_Own_percent_rank'] = portfolio.iloc[:,1].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX2_Own_percent_rank'] = portfolio.iloc[:,2].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX3_Own_percent_rank'] = portfolio.iloc[:,3].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX4_Own_percent_rank'] = portfolio.iloc[:,4].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
            portfolio['FLEX5_Own_percent_rank'] = portfolio.iloc[:,5].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])

            portfolio['CPT_Own'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']) / 100
            portfolio['FLEX1_Own'] = portfolio.iloc[:,1].map(maps_dict['own_map']) / 100
            portfolio['FLEX2_Own'] = portfolio.iloc[:,2].map(maps_dict['own_map']) / 100
            portfolio['FLEX3_Own'] = portfolio.iloc[:,3].map(maps_dict['own_map']) / 100
            portfolio['FLEX4_Own'] = portfolio.iloc[:,4].map(maps_dict['own_map']) / 100
            portfolio['FLEX5_Own'] = portfolio.iloc[:,5].map(maps_dict['own_map']) / 100

            portfolio['own_product'] = (portfolio[own_columns].product(axis=1))
            portfolio['own_average'] = (portfolio['Own'].max() * .33) / 100
            portfolio['own_sum'] = portfolio[own_columns].sum(axis=1)
            portfolio['avg_own_rank'] = portfolio[dup_count_columns].mean(axis=1)
            
            # Calculate dupes formula
            portfolio['dupes_calc'] = (portfolio['own_product'] * portfolio['avg_own_rank']) * Contest_Size + ((portfolio['salary'] - (50000 - portfolio['Own'])) / 100) - ((50000 - portfolio['salary']) / 100)
            portfolio['dupes_calc'] = portfolio['dupes_calc'] * dupes_multiplier

            # Round and handle negative values
            portfolio['Dupes'] = np.where(
                np.round(portfolio['dupes_calc'], 0) <= 0,
                0, 
                np.round(portfolio['dupes_calc'], 0) - 1
            )
        elif type_var == 'Classic':
            if sport_var == 'CS2':
                dup_count_columns = ['CPT_Own_percent_rank', 'FLEX1_Own_percent_rank', 'FLEX2_Own_percent_rank', 'FLEX3_Own_percent_rank', 'FLEX4_Own_percent_rank', 'FLEX5_Own_percent_rank']
                own_columns = ['CPT_Own', 'FLEX1_Own', 'FLEX2_Own', 'FLEX3_Own', 'FLEX4_Own', 'FLEX5_Own']
                calc_columns = ['own_product', 'own_average', 'own_sum', 'avg_own_rank', 'dupes_calc', 'low_own_count', 'Ref_Proj', 'Max_Proj', 'Min_Proj', 'Avg_Ref', 'own_ratio']
                # Get the original player columns (first 6 columns excluding salary, median, Own)
                player_columns = [col for col in portfolio.columns[:6] if col not in ['salary', 'median', 'Own']]
                
                flex_ownerships = pd.concat([
                    portfolio.iloc[:,1].map(maps_dict['own_map']),
                    portfolio.iloc[:,2].map(maps_dict['own_map']),
                    portfolio.iloc[:,3].map(maps_dict['own_map']),
                    portfolio.iloc[:,4].map(maps_dict['own_map']),
                    portfolio.iloc[:,5].map(maps_dict['own_map'])
                ])
                flex_rank = flex_ownerships.rank(pct=True)
                
                # Assign ranks back to individual columns using the same rank scale
                portfolio['CPT_Own_percent_rank'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']).rank(pct=True)
                portfolio['FLEX1_Own_percent_rank'] = portfolio.iloc[:,1].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
                portfolio['FLEX2_Own_percent_rank'] = portfolio.iloc[:,2].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
                portfolio['FLEX3_Own_percent_rank'] = portfolio.iloc[:,3].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
                portfolio['FLEX4_Own_percent_rank'] = portfolio.iloc[:,4].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])
                portfolio['FLEX5_Own_percent_rank'] = portfolio.iloc[:,5].map(maps_dict['own_map']).map(lambda x: flex_rank[flex_ownerships == x].iloc[0])

                portfolio['CPT_Own'] = portfolio.iloc[:,0].map(maps_dict['cpt_own_map']) / 100
                portfolio['FLEX1_Own'] = portfolio.iloc[:,1].map(maps_dict['own_map']) / 100
                portfolio['FLEX2_Own'] = portfolio.iloc[:,2].map(maps_dict['own_map']) / 100
                portfolio['FLEX3_Own'] = portfolio.iloc[:,3].map(maps_dict['own_map']) / 100
                portfolio['FLEX4_Own'] = portfolio.iloc[:,4].map(maps_dict['own_map']) / 100
                portfolio['FLEX5_Own'] = portfolio.iloc[:,5].map(maps_dict['own_map']) / 100

                portfolio['own_product'] = (portfolio[own_columns].product(axis=1))
                portfolio['own_average'] = (portfolio['Own'].max() * .33) / 100
                portfolio['own_sum'] = portfolio[own_columns].sum(axis=1)
                portfolio['avg_own_rank'] = portfolio[dup_count_columns].mean(axis=1)
                
                # Calculate dupes formula
                portfolio['dupes_calc'] = (portfolio['own_product'] * portfolio['avg_own_rank']) * Contest_Size + ((portfolio['salary'] - (50000 - portfolio['Own'])) / 100) - ((50000 - portfolio['salary']) / 100)
                portfolio['dupes_calc'] = portfolio['dupes_calc'] * dupes_multiplier

                # Round and handle negative values
                portfolio['Dupes'] = np.where(
                    np.round(portfolio['dupes_calc'], 0) <= 0,
                    0, 
                    np.round(portfolio['dupes_calc'], 0) - 1
                )
            elif sport_var != 'CS2':
                num_players = len([col for col in portfolio.columns if col not in ['salary', 'median', 'Own']])
                dup_count_columns = [f'player_{i}_percent_rank' for i in range(1, num_players + 1)]
                own_columns = [f'player_{i}_own' for i in range(1, num_players + 1)]
                calc_columns = ['own_product', 'own_average', 'own_sum', 'avg_own_rank', 'dupes_calc', 'low_own_count', 'Ref_Proj', 'Max_Proj', 'Min_Proj', 'Avg_Ref', 'own_ratio']
                # Get the original player columns (first num_players columns excluding salary, median, Own)
                player_columns = [col for col in portfolio.columns[:num_players] if col not in ['salary', 'median', 'Own']]
                
                for i in range(1, num_players + 1):
                    portfolio[f'player_{i}_percent_rank'] = portfolio.iloc[:,i-1].map(maps_dict['own_percent_rank'])
                    portfolio[f'player_{i}_own'] = portfolio.iloc[:,i-1].map(maps_dict['own_map']) / 100
                
                portfolio['own_product'] = (portfolio[own_columns].product(axis=1))
                portfolio['own_average'] = (portfolio['Own'].max() * .33) / 100
                portfolio['own_sum'] = portfolio[own_columns].sum(axis=1)
                portfolio['avg_own_rank'] = portfolio[dup_count_columns].mean(axis=1)
                
                portfolio['dupes_calc'] = (portfolio['own_product'] * portfolio['avg_own_rank']) * Contest_Size + ((portfolio['salary'] - (50000 - portfolio['Own'])) / 100) - ((50000 - portfolio['salary']) / 100)
                portfolio['dupes_calc'] = portfolio['dupes_calc'] * dupes_multiplier
                # Round and handle negative values
                portfolio['Dupes'] = np.where(
                    np.round(portfolio['dupes_calc'], 0) <= 0,
                    0, 
                    np.round(portfolio['dupes_calc'], 0) - 1
                )

    portfolio['Dupes'] = np.round(portfolio['Dupes'], 0)
    portfolio['own_ratio'] = np.where(
        portfolio[own_columns].isin([max_ownership]).any(axis=1),
        portfolio['own_sum'] / portfolio['own_average'],
        (portfolio['own_sum'] - max_ownership) / portfolio['own_average']
    )
    percentile_cut_scalar = portfolio['median'].max()  # Get scalar value
    if type_var == 'Classic':
        if sport_var == 'CS2':
            own_ratio_nerf = 2
        elif sport_var != 'CS2':
            own_ratio_nerf = 1.5
    elif type_var == 'Showdown':
        own_ratio_nerf = 1.5
    portfolio['Finish_percentile'] = portfolio.apply(
        lambda row: .0005 if (row['own_ratio'] - own_ratio_nerf) / ((10 * (row['median'] / percentile_cut_scalar)) / 2) < .0005 
        else (row['own_ratio'] - own_ratio_nerf) / ((10 * (row['median'] / percentile_cut_scalar)) / 2), 
        axis=1
    )
    
    portfolio['Ref_Proj'] = portfolio['median'].max()
    portfolio['Max_Proj'] = portfolio['Ref_Proj'] + 10
    portfolio['Min_Proj'] = portfolio['Ref_Proj'] - 10
    portfolio['Avg_Ref'] = (portfolio['Max_Proj'] + portfolio['Min_Proj']) / 2
    portfolio['Win%'] = (((portfolio['median'] / portfolio['Avg_Ref']) - (0.1 + ((portfolio['Ref_Proj'] - portfolio['median'])/100))) / (Contest_Size / 1000)) / 10
    max_allowed_win = (1 / Contest_Size) * 5
    portfolio['Win%'] = portfolio['Win%'] / portfolio['Win%'].max() * max_allowed_win
    
    portfolio['Finish_percentile'] = portfolio['Finish_percentile'] + .005 + (.005 * (Contest_Size / 10000))
    portfolio['Finish_percentile'] = portfolio['Finish_percentile'] * percentile_multiplier
    portfolio['Win%'] = portfolio['Win%'] * (1 - portfolio['Finish_percentile'])
    
    portfolio['low_own_count'] = portfolio[own_columns].apply(lambda row: (row < 0.10).sum(), axis=1)
    portfolio['Finish_percentile'] = portfolio.apply(lambda row: row['Finish_percentile'] if row['low_own_count'] <= 0 else row['Finish_percentile'] / row['low_own_count'], axis=1)
    portfolio['Lineup Edge'] = portfolio['Win%'] * ((.5 - portfolio['Finish_percentile']) * (Contest_Size / 2.5))
    portfolio['Lineup Edge'] = portfolio.apply(lambda row: row['Lineup Edge'] / (row['Dupes'] + 1) if row['Dupes'] > 0 else row['Lineup Edge'], axis=1)
    portfolio['Lineup Edge'] = portfolio['Lineup Edge'] - portfolio['Lineup Edge'].mean()
    portfolio['Weighted Own'] = portfolio[own_columns].apply(calculate_weighted_ownership, axis=1)
    portfolio['Geomean'] = np.power((portfolio[own_columns] * 100).product(axis=1), 1 / len(own_columns))
    
    # Calculate similarity score based on actual player selection
    portfolio['Similarity Score'] = calculate_player_similarity_score(portfolio, player_columns)
    
    portfolio = portfolio.drop(columns=dup_count_columns)
    portfolio = portfolio.drop(columns=own_columns)
    portfolio = portfolio.drop(columns=calc_columns)

    return portfolio