app.py

import streamlit as st
st.set_page_config(layout="wide")
import numpy as np
import pandas as pd
import time
from fuzzywuzzy import process
import random

## import global functions
from global_func.clean_player_name import clean_player_name
from global_func.load_file import load_file
from global_func.load_ss_file import load_ss_file
from global_func.find_name_mismatches import find_name_mismatches
from global_func.predict_dupes import predict_dupes
from global_func.highlight_rows import highlight_changes, highlight_changes_winners, highlight_changes_losers
from global_func.load_csv import load_csv
from global_func.find_csv_mismatches import find_csv_mismatches

freq_format = {'Finish_percentile': '{:.2%}', 'Lineup Edge': '{:.2%}', 'Win%': '{:.2%}'}
player_wrong_names_mlb = ['Enrique Hernandez']
player_right_names_mlb = ['Kike Hernandez']

tab1, tab2, tab3 = st.tabs(["Data Load", "Late Swap", "Manage Portfolio"])
with tab1:
    if st.button('Clear data', key='reset1'):
        st.session_state.clear()
    # Add file uploaders to your app
    col1, col2, col3 = st.columns(3)

    with col1:
        st.subheader("Draftkings/Fanduel CSV")
        st.info("Upload the player pricing CSV from the site you are playing on. This is used in late swap exporting and/or with SaberSim portfolios, but is not necessary for the portfolio management functions.")

        upload_csv_col, csv_template_col = st.columns([3, 1])
        with upload_csv_col:
            csv_file = st.file_uploader("Upload CSV File", type=['csv'])
            if 'csv_file' in st.session_state:
                del st.session_state['csv_file']
        with csv_template_col:

            csv_template_df = pd.DataFrame(columns=['Name', 'ID', 'Roster Position', 'Salary'])

            st.download_button(
                label="CSV Template",
                data=csv_template_df.to_csv(index=False),
                file_name="csv_template.csv",
                mime="text/csv"
            )
        st.session_state['csv_file'] = load_csv(csv_file)
        try:
            st.session_state['csv_file']['Salary'] = st.session_state['csv_file']['Salary'].astype(str).str.replace(',', '').astype(int)
        except:
            pass
            
        if csv_file:
            st.session_state['csv_file'] = st.session_state['csv_file'].drop_duplicates(subset=['Name'])
            st.success('Projections file loaded successfully!')
            st.dataframe(st.session_state['csv_file'].head(10))
    
    with col2:
        st.subheader("Portfolio File")
        st.info("Go ahead and upload a portfolio file here. Only include player columns and an optional 'Stack' column if you are playing MLB.")
        saber_toggle = st.radio("Are you uploading from SaberSim?", options=['No', 'Yes'])
        st.info("If you are uploading from SaberSim, you will need to upload a CSV file for the slate for name matching.")
        if saber_toggle == 'Yes':
            if csv_file is not None:
                portfolio_file = st.file_uploader("Upload Portfolio File (CSV or Excel)", type=['csv', 'xlsx', 'xls'])
                if 'portfolio' in st.session_state:
                    del st.session_state['portfolio']
                if 'export_portfolio' in st.session_state:
                    del st.session_state['export_portfolio']

        else:
            portfolio_file = st.file_uploader("Upload Portfolio File (CSV or Excel)", type=['csv', 'xlsx', 'xls'])
            if 'portfolio' in st.session_state:
                del st.session_state['portfolio']
            if 'export_portfolio' in st.session_state:
                del st.session_state['export_portfolio']

        if portfolio_file:
            if saber_toggle == 'Yes':
                st.session_state['export_portfolio'], st.session_state['portfolio'] = load_ss_file(portfolio_file, st.session_state['csv_file'])
                st.session_state['export_portfolio'] = st.session_state['export_portfolio'].dropna(how='all')
                st.session_state['export_portfolio'] = st.session_state['export_portfolio'].reset_index(drop=True)
                st.session_state['portfolio'] = st.session_state['portfolio'].dropna(how='all')
                st.session_state['portfolio'] = st.session_state['portfolio'].reset_index(drop=True)
            else:
                st.session_state['export_portfolio'], st.session_state['portfolio'] = load_file(portfolio_file)
                st.session_state['export_portfolio'] = st.session_state['export_portfolio'].dropna(how='all')
                st.session_state['export_portfolio'] = st.session_state['export_portfolio'].reset_index(drop=True)
                st.session_state['portfolio'] = st.session_state['portfolio'].dropna(how='all')
                st.session_state['portfolio'] = st.session_state['portfolio'].reset_index(drop=True)
            # Check if Stack column exists in the portfolio
            if 'Stack' in st.session_state['portfolio'].columns:
                # Create dictionary mapping index to Stack values
                stack_dict = dict(zip(st.session_state['portfolio'].index, st.session_state['portfolio']['Stack']))
                st.write(f"Found {len(stack_dict)} stack assignments")
                st.session_state['portfolio'] = st.session_state['portfolio'].drop(columns=['Stack'])
            else:
                stack_dict = None
                st.info("No Stack column found in portfolio")
            if st.session_state['portfolio'] is not None:
                st.success('Portfolio file loaded successfully!')
                st.session_state['portfolio'] = st.session_state['portfolio'].apply(lambda x: x.replace(player_wrong_names_mlb, player_right_names_mlb))
                st.dataframe(st.session_state['portfolio'].head(10))

    with col3:
        st.subheader("Projections File")
        st.info("upload a projections file that has 'player_names', 'salary', 'median', 'ownership', and 'captain ownership' (Needed for Showdown) columns. Note that the salary for showdown needs to be the FLEX salary, not the captain salary.")
        
        # Create two columns for the uploader and template button
        upload_col, template_col = st.columns([3, 1])
        
        with upload_col:
            projections_file = st.file_uploader("Upload Projections File (CSV or Excel)", type=['csv', 'xlsx', 'xls'])
            if 'projections_df' in st.session_state:
                del st.session_state['projections_df']
        
        with template_col:
            # Create empty DataFrame with required columns
            template_df = pd.DataFrame(columns=['player_names', 'position', 'team', 'salary', 'median', 'ownership', 'captain ownership'])
            # Add download button for template
            st.download_button(
                label="Template",
                data=template_df.to_csv(index=False),
                file_name="projections_template.csv",
                mime="text/csv"
            )
            
        if projections_file:
            export_projections, projections = load_file(projections_file)
            if projections is not None:
                st.success('Projections file loaded successfully!')
                projections = projections.apply(lambda x: x.replace(player_wrong_names_mlb, player_right_names_mlb))
                st.dataframe(projections.head(10))

    if portfolio_file and projections_file:
        if st.session_state['portfolio'] is not None and projections is not None:
            st.subheader("Name Matching Analysis")
            # Initialize projections_df in session state if it doesn't exist
            if 'projections_df' not in st.session_state:
                st.session_state['projections_df'] = projections.copy()
                st.session_state['projections_df']['salary'] = (st.session_state['projections_df']['salary'].astype(str).str.replace(',', '').astype(float).astype(int))
            
            # Update projections_df with any new matches
            st.session_state['projections_df'] = find_name_mismatches(st.session_state['portfolio'], st.session_state['projections_df'])
            if csv_file is not None and 'export_dict' not in st.session_state:
                    # Create a dictionary of Name to Name+ID from csv_file
                    try:
                        name_id_map = dict(zip(
                            st.session_state['csv_file']['Name'], 
                            st.session_state['csv_file']['Name + ID']
                        ))
                    except:
                        name_id_map = dict(zip(
                            st.session_state['csv_file']['Nickname'], 
                            st.session_state['csv_file']['Id']
                        ))
                    
                    # Function to find best match
                    def find_best_match(name):
                        best_match = process.extractOne(name, name_id_map.keys())
                        if best_match and best_match[1] >= 85:  # 85% match threshold
                            return name_id_map[best_match[0]]
                        return name  # Return original name if no good match found
                    
                    # Apply the matching
                    projections['upload_match'] = projections['player_names'].apply(find_best_match)
                    st.session_state['export_dict'] = dict(zip(projections['player_names'], projections['upload_match']))

with tab2:
    if st.button('Clear data', key='reset2'):
        st.session_state.clear()
    
    if 'portfolio' in st.session_state and 'projections_df' in st.session_state:

        optimized_df = None

        map_dict = {
                    'pos_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                    st.session_state['projections_df']['position'])),
                    'salary_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['salary'])),
                    'proj_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                    st.session_state['projections_df']['median'])),
                    'own_map': dict(zip(st.session_state['projections_df']['player_names'],
                                    st.session_state['projections_df']['ownership'])),
                    'team_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['team']))
                }
        # Calculate new stats for optimized lineups
        st.session_state['portfolio']['salary'] = st.session_state['portfolio'].apply(
            lambda row: sum(map_dict['salary_map'].get(player, 0) for player in row if player in map_dict['salary_map']), axis=1
        )
        st.session_state['portfolio']['median'] = st.session_state['portfolio'].apply(
            lambda row: sum(map_dict['proj_map'].get(player, 0) for player in row if player in map_dict['proj_map']), axis=1
        )

        st.session_state['portfolio']['Own'] = st.session_state['portfolio'].apply(
            lambda row: sum(map_dict['own_map'].get(player, 0) for player in row if player in map_dict['own_map']), axis=1
        )

        options_container = st.container()
        with options_container:
            col1, col2, col3, col4, col5, col6 = st.columns(6)
            with col1:
                curr_site_var = st.selectbox("Select your current site", options=['DraftKings', 'FanDuel'])
            with col2:
                curr_sport_var = st.selectbox("Select your current sport", options=['NBA', 'MLB', 'NFL', 'NHL', 'MMA'])
            with col3:
                swap_var = st.multiselect("Select late swap strategy", options=['Optimize', 'Increase volatility', 'Decrease volatility'])
            with col4:
                remove_teams_var = st.multiselect("What teams have already played?", options=st.session_state['projections_df']['team'].unique())
            with col5:
                winners_var = st.multiselect("Are there any players doing exceptionally well?", options=st.session_state['projections_df']['player_names'].unique(), max_selections=3)
            with col6:
                losers_var = st.multiselect("Are there any players doing exceptionally poorly?", options=st.session_state['projections_df']['player_names'].unique(), max_selections=3)
        if st.button('Clear Late Swap'):
            if 'optimized_df' in st.session_state:
                del st.session_state['optimized_df']

            map_dict = {
                        'pos_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['position'])),
                        'salary_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                            st.session_state['projections_df']['salary'])),
                        'proj_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['median'])),
                        'own_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['ownership'])),
                        'team_map': dict(zip(st.session_state['projections_df']['player_names'], 
                                        st.session_state['projections_df']['team']))
                    }
            # Calculate new stats for optimized lineups
            st.session_state['portfolio']['salary'] = st.session_state['portfolio'].apply(
                lambda row: sum(map_dict['salary_map'].get(player, 0) for player in row if player in map_dict['salary_map']), axis=1
            )
            st.session_state['portfolio']['median'] = st.session_state['portfolio'].apply(
                lambda row: sum(map_dict['proj_map'].get(player, 0) for player in row if player in map_dict['proj_map']), axis=1
            )
            st.session_state['portfolio']['Own'] = st.session_state['portfolio'].apply(
                lambda row: sum(map_dict['own_map'].get(player, 0) for player in row if player in map_dict['own_map']), axis=1
            )

        if st.button('Run Late Swap'):
            st.session_state['portfolio'] = st.session_state['portfolio'].drop(columns=['salary', 'median', 'Own'])
            if curr_sport_var == 'NBA':
                if curr_site_var == 'DraftKings':
                    st.session_state['portfolio'] = st.session_state['portfolio'].set_axis(['PG', 'SG', 'SF', 'PF', 'C', 'G', 'F', 'UTIL'], axis=1)
                else:
                    st.session_state['portfolio'] = st.session_state['portfolio'].set_axis(['PG', 'PG', 'SG', 'SG', 'SF', 'SF', 'PF', 'PF', 'C'], axis=1)
            
            # Define roster position rules
            if curr_site_var == 'DraftKings':
                position_rules = {
                    'PG': ['PG'],
                    'SG': ['SG'],
                    'SF': ['SF'],
                    'PF': ['PF'],
                    'C': ['C'],
                    'G': ['PG', 'SG'],
                    'F': ['SF', 'PF'],
                    'UTIL': ['PG', 'SG', 'SF', 'PF', 'C']
                    }
            else:
                position_rules = {
                    'PG': ['PG'],
                    'SG': ['SG'],
                    'SF': ['SF'],
                    'PF': ['PF'],
                    'C': ['C'],
                }
            # Create position groups from projections data
            position_groups = {}
            for _, player in st.session_state['projections_df'].iterrows():
                positions = player['position'].split('/')
                for pos in positions:
                    if pos not in position_groups:
                        position_groups[pos] = []
                    position_groups[pos].append({
                        'player_names': player['player_names'],
                        'salary': player['salary'],
                        'median': player['median'],
                        'ownership': player['ownership'],
                        'positions': positions  # Store all eligible positions
                    })

            def optimize_lineup(row):
                current_lineup = []
                total_salary = 0
                if curr_site_var == 'DraftKings':
                    salary_cap = 50000
                else:
                    salary_cap = 60000
                used_players = set()

                # Convert row to dictionary with roster positions
                roster = {}
                for col, player in zip(row.index, row):
                    if col not in ['salary', 'median', 'Own', 'Finish_percentile', 'Dupes', 'Lineup Edge']:
                        roster[col] = {
                            'name': player,
                            'position': map_dict['pos_map'].get(player, '').split('/'),
                            'team': map_dict['team_map'].get(player, ''),
                            'salary': map_dict['salary_map'].get(player, 0),
                            'median': map_dict['proj_map'].get(player, 0),
                            'ownership': map_dict['own_map'].get(player, 0)
                        }
                        total_salary += roster[col]['salary']
                        used_players.add(player)

                # Optimize each roster position in random order
                roster_positions = list(roster.items())
                random.shuffle(roster_positions)
                
                for roster_pos, current in roster_positions:
                    # Skip optimization for players from removed teams
                    if current['team'] in remove_teams_var:
                        continue
                    
                    valid_positions = position_rules[roster_pos]
                    better_options = []

                    # Find valid replacements for this roster position
                    for pos in valid_positions:
                        if pos in position_groups:
                            pos_options = [
                                p for p in position_groups[pos]
                                if p['median'] > current['median']
                                and (total_salary - current['salary'] + p['salary']) <= salary_cap
                                and p['player_names'] not in used_players
                                and any(valid_pos in p['positions'] for valid_pos in valid_positions)
                                and map_dict['team_map'].get(p['player_names']) not in remove_teams_var  # Check team restriction
                            ]
                            better_options.extend(pos_options)

                    if better_options:
                        # Remove duplicates
                        better_options = {opt['player_names']: opt for opt in better_options}.values()
                        
                        # Sort by median projection and take the best one
                        best_replacement = max(better_options, key=lambda x: x['median'])
                        
                        # Update the lineup and tracking variables
                        used_players.remove(current['name'])
                        used_players.add(best_replacement['player_names'])
                        total_salary = total_salary - current['salary'] + best_replacement['salary']
                        roster[roster_pos] = {
                            'name': best_replacement['player_names'],
                            'position': map_dict['pos_map'][best_replacement['player_names']].split('/'),
                            'team': map_dict['team_map'][best_replacement['player_names']],
                            'salary': best_replacement['salary'],
                            'median': best_replacement['median'],
                            'ownership': best_replacement['ownership']
                        }

                # Return optimized lineup maintaining original column order
                return [roster[pos]['name'] for pos in row.index if pos in roster]

            def optimize_lineup_winners(row):
                current_lineup = []
                total_salary = 0
                if curr_site_var == 'DraftKings':
                    salary_cap = 50000
                else:
                    salary_cap = 60000
                used_players = set()

                # Check if any winners are in the lineup and count them
                winners_in_lineup = sum(1 for player in row if player in winners_var)
                changes_needed = min(winners_in_lineup, 3) if winners_in_lineup > 0 else 0
                changes_made = 0

                # Convert row to dictionary with roster positions
                roster = {}
                for col, player in zip(row.index, row):
                    if col not in ['salary', 'median', 'Own', 'Finish_percentile', 'Dupes', 'Lineup Edge']:
                        roster[col] = {
                            'name': player,
                            'position': map_dict['pos_map'].get(player, '').split('/'),
                            'team': map_dict['team_map'].get(player, ''),
                            'salary': map_dict['salary_map'].get(player, 0),
                            'median': map_dict['proj_map'].get(player, 0),
                            'ownership': map_dict['own_map'].get(player, 0)
                        }
                        total_salary += roster[col]['salary']
                        used_players.add(player)

                # Only proceed with ownership-based optimization if we have winners in the lineup
                if changes_needed > 0:
                    # Randomize the order of positions to optimize
                    roster_positions = list(roster.items())
                    random.shuffle(roster_positions)
                    
                    for roster_pos, current in roster_positions:
                        # Stop if we've made enough changes
                        if changes_made >= changes_needed:
                            break
                            
                        # Skip optimization for players from removed teams or if the current player is a winner
                        if current['team'] in remove_teams_var or current['name'] in winners_var:
                            continue
                        
                        valid_positions = list(position_rules[roster_pos])
                        random.shuffle(valid_positions)
                        better_options = []

                        # Find valid replacements with higher ownership
                        for pos in valid_positions:
                            if pos in position_groups:
                                pos_options = [
                                    p for p in position_groups[pos]
                                    if p['ownership'] > current['ownership']
                                    and p['median'] >= current['median'] - 3
                                    and (total_salary - current['salary'] + p['salary']) <= salary_cap
                                    and (total_salary - current['salary'] + p['salary']) >= salary_cap - 1000
                                    and p['player_names'] not in used_players
                                    and any(valid_pos in p['positions'] for valid_pos in valid_positions)
                                    and map_dict['team_map'].get(p['player_names']) not in remove_teams_var
                                ]
                                better_options.extend(pos_options)

                        if better_options:
                            # Remove duplicates
                            better_options = {opt['player_names']: opt for opt in better_options}.values()
                            
                            # Sort by ownership and take the highest owned option
                            best_replacement = max(better_options, key=lambda x: x['ownership'])
                            
                            # Update the lineup and tracking variables
                            used_players.remove(current['name'])
                            used_players.add(best_replacement['player_names'])
                            total_salary = total_salary - current['salary'] + best_replacement['salary']
                            roster[roster_pos] = {
                                'name': best_replacement['player_names'],
                                'position': map_dict['pos_map'][best_replacement['player_names']].split('/'),
                                'team': map_dict['team_map'][best_replacement['player_names']],
                                'salary': best_replacement['salary'],
                                'median': best_replacement['median'],
                                'ownership': best_replacement['ownership']
                            }
                            changes_made += 1

                # Return optimized lineup maintaining original column order
                return [roster[pos]['name'] for pos in row.index if pos in roster]
            
            def optimize_lineup_losers(row):
                current_lineup = []
                total_salary = 0
                if curr_site_var == 'DraftKings':
                    salary_cap = 50000
                else:
                    salary_cap = 60000
                used_players = set()

                # Check if any winners are in the lineup and count them
                losers_in_lineup = sum(1 for player in row if player in losers_var)
                changes_needed = min(losers_in_lineup, 3) if losers_in_lineup > 0 else 0
                changes_made = 0

                # Convert row to dictionary with roster positions
                roster = {}
                for col, player in zip(row.index, row):
                    if col not in ['salary', 'median', 'Own', 'Finish_percentile', 'Dupes', 'Lineup Edge']:
                        roster[col] = {
                            'name': player,
                            'position': map_dict['pos_map'].get(player, '').split('/'),
                            'team': map_dict['team_map'].get(player, ''),
                            'salary': map_dict['salary_map'].get(player, 0),
                            'median': map_dict['proj_map'].get(player, 0),
                            'ownership': map_dict['own_map'].get(player, 0)
                        }
                        total_salary += roster[col]['salary']
                        used_players.add(player)

                # Only proceed with ownership-based optimization if we have winners in the lineup
                if changes_needed > 0:
                    # Randomize the order of positions to optimize
                    roster_positions = list(roster.items())
                    random.shuffle(roster_positions)
                    
                    for roster_pos, current in roster_positions:
                        # Stop if we've made enough changes
                        if changes_made >= changes_needed:
                            break
                            
                        # Skip optimization for players from removed teams or if the current player is a winner
                        if current['team'] in remove_teams_var or current['name'] in losers_var:
                            continue
                        
                        valid_positions = list(position_rules[roster_pos])
                        random.shuffle(valid_positions)
                        better_options = []

                        # Find valid replacements with higher ownership
                        for pos in valid_positions:
                            if pos in position_groups:
                                pos_options = [
                                    p for p in position_groups[pos]
                                    if p['ownership'] < current['ownership']
                                    and p['median'] >= current['median'] - 3
                                    and (total_salary - current['salary'] + p['salary']) <= salary_cap
                                    and (total_salary - current['salary'] + p['salary']) >= salary_cap - 1000
                                    and p['player_names'] not in used_players
                                    and any(valid_pos in p['positions'] for valid_pos in valid_positions)
                                    and map_dict['team_map'].get(p['player_names']) not in remove_teams_var
                                ]
                                better_options.extend(pos_options)

                        if better_options:
                            # Remove duplicates
                            better_options = {opt['player_names']: opt for opt in better_options}.values()
                            
                            # Sort by ownership and take the highest owned option
                            best_replacement = max(better_options, key=lambda x: x['ownership'])
                            
                            # Update the lineup and tracking variables
                            used_players.remove(current['name'])
                            used_players.add(best_replacement['player_names'])
                            total_salary = total_salary - current['salary'] + best_replacement['salary']
                            roster[roster_pos] = {
                                'name': best_replacement['player_names'],
                                'position': map_dict['pos_map'][best_replacement['player_names']].split('/'),
                                'team': map_dict['team_map'][best_replacement['player_names']],
                                'salary': best_replacement['salary'],
                                'median': best_replacement['median'],
                                'ownership': best_replacement['ownership']
                            }
                            changes_made += 1

                # Return optimized lineup maintaining original column order
                return [roster[pos]['name'] for pos in row.index if pos in roster]

            # Create a progress bar
            progress_bar = st.progress(0)
            status_text = st.empty()
            
            # Process each lineup
            optimized_lineups = []
            total_lineups = len(st.session_state['portfolio'])
            
            for idx, row in st.session_state['portfolio'].iterrows():
                # First optimization pass
                first_pass = optimize_lineup(row)
                first_pass_series = pd.Series(first_pass, index=row.index)

                second_pass = optimize_lineup(first_pass_series)
                second_pass_series = pd.Series(second_pass, index=row.index)

                third_pass = optimize_lineup(second_pass_series)
                third_pass_series = pd.Series(third_pass, index=row.index)

                fourth_pass = optimize_lineup(third_pass_series)
                fourth_pass_series = pd.Series(fourth_pass, index=row.index)

                fifth_pass = optimize_lineup(fourth_pass_series)
                fifth_pass_series = pd.Series(fifth_pass, index=row.index)
                
                # Second optimization pass
                final_lineup = optimize_lineup(fifth_pass_series)
                optimized_lineups.append(final_lineup)
                
                if 'Optimize' in swap_var:
                    progress = (idx + 1) / total_lineups
                    progress_bar.progress(progress)
                    status_text.text(f'Optimizing Lineups {idx + 1} of {total_lineups}')
                else:
                    pass
            
            # Create new dataframe with optimized lineups
            if 'Optimize' in swap_var:
                st.session_state['optimized_df_medians'] = pd.DataFrame(optimized_lineups, columns=st.session_state['portfolio'].columns)
            else:
                st.session_state['optimized_df_medians'] = st.session_state['portfolio']

            # Create a progress bar
            progress_bar_winners = st.progress(0)
            status_text_winners = st.empty()
            
            # Process each lineup
            optimized_lineups_winners = []
            total_lineups = len(st.session_state['optimized_df_medians'])
            
            for idx, row in st.session_state['optimized_df_medians'].iterrows():

                final_lineup = optimize_lineup_winners(row)
                optimized_lineups_winners.append(final_lineup)
                
                if 'Decrease volatility' in swap_var:
                    progress_winners = (idx + 1) / total_lineups
                    progress_bar_winners.progress(progress_winners)
                    status_text_winners.text(f'Lowering Volatility around Winners {idx + 1} of {total_lineups}')
                else:
                    pass
            
            # Create new dataframe with optimized lineups
            if 'Decrease volatility' in swap_var:
                st.session_state['optimized_df_winners'] = pd.DataFrame(optimized_lineups_winners, columns=st.session_state['optimized_df_medians'].columns)
            else:
                st.session_state['optimized_df_winners'] = st.session_state['optimized_df_medians']

            # Create a progress bar
            progress_bar_losers = st.progress(0)
            status_text_losers = st.empty()
            
            # Process each lineup
            optimized_lineups_losers = []
            total_lineups = len(st.session_state['optimized_df_winners'])
            
            for idx, row in st.session_state['optimized_df_winners'].iterrows():

                final_lineup = optimize_lineup_losers(row)
                optimized_lineups_losers.append(final_lineup)
                
                if 'Increase volatility' in swap_var:
                    progress_losers = (idx + 1) / total_lineups
                    progress_bar_losers.progress(progress_losers)
                    status_text_losers.text(f'Increasing Volatility around Losers {idx + 1} of {total_lineups}')
                else:
                    pass
            
            # Create new dataframe with optimized lineups
            if 'Increase volatility' in swap_var:
                st.session_state['optimized_df'] = pd.DataFrame(optimized_lineups_losers, columns=st.session_state['optimized_df_winners'].columns)
            else:
                st.session_state['optimized_df'] = st.session_state['optimized_df_winners']
            
            # Calculate new stats for optimized lineups
            st.session_state['optimized_df']['salary'] = st.session_state['optimized_df'].apply(
                lambda row: sum(map_dict['salary_map'].get(player, 0) for player in row if player in map_dict['salary_map']), axis=1
            )
            st.session_state['optimized_df']['median'] = st.session_state['optimized_df'].apply(
                lambda row: sum(map_dict['proj_map'].get(player, 0) for player in row if player in map_dict['proj_map']), axis=1
            )
            st.session_state['optimized_df']['Own'] = st.session_state['optimized_df'].apply(
                lambda row: sum(map_dict['own_map'].get(player, 0) for player in row if player in map_dict['own_map']), axis=1
            )

            # Display results
            st.success('Optimization complete!')

        if 'optimized_df' in st.session_state:
            st.write("Increase in median highlighted in yellow, descrease in volatility highlighted in blue, increase in volatility highlighted in red:")
            st.dataframe(
                st.session_state['optimized_df'].style
                .apply(highlight_changes, axis=1)
                .apply(highlight_changes_winners, axis=1)
                .apply(highlight_changes_losers, axis=1)
                .background_gradient(axis=0)
                .background_gradient(cmap='RdYlGn')
                .format(precision=2),
                height=1000,
                use_container_width=True
            )
            
            # Option to download optimized lineups
            if st.button('Prepare Late Swap Export'):
                export_df = st.session_state['optimized_df'].copy()
                
                # Map player names to their export IDs for all player columns
                for col in export_df.columns:
                    if col not in ['salary', 'median', 'Own']:
                        export_df[col] = export_df[col].map(st.session_state['export_dict'])
                
                csv = export_df.to_csv(index=False)
                st.download_button(
                    label="Download CSV",
                    data=csv,
                    file_name="optimized_lineups.csv",
                    mime="text/csv"
                )
        else:
            st.write("Current Portfolio")
            st.dataframe(
                st.session_state['portfolio'].style
                .background_gradient(axis=0)
                .background_gradient(cmap='RdYlGn')
                .format(precision=2),
                height=1000,
                use_container_width=True
            )

with tab3:
    if st.button('Clear data', key='reset3'):
        st.session_state.clear()
    if 'portfolio' in st.session_state and 'projections_df' in st.session_state:
        col1, col2, col3 = st.columns([1, 8, 1])
        excluded_cols = ['salary', 'median', 'Own', 'Finish_percentile', 'Dupes', 'Stack', 'Win%', 'Lineup Edge']
        with col1:
            site_var = st.selectbox("Select Site", ['Draftkings', 'Fanduel'])
            sport_var = st.selectbox("Select Sport", ['NFL', 'MLB', 'NBA', 'NHL', 'MMA'])
            st.info("It currently does not matter what sport you select, it may matter in the future.")
            type_var = st.selectbox("Select Game Type", ['Classic', 'Showdown'])
            Contest_Size = st.number_input("Enter Contest Size", value=25000, min_value=1, step=1)
            strength_var = st.selectbox("Select field strength", ['Average', 'Sharp', 'Weak'])
        if site_var == 'Draftkings':
            if type_var == 'Classic':
                map_dict = {
                    'pos_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['position'])),
                    'team_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['team'])),
                    'salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                    'proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'])),
                    'own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'])),
                    'own_percent_rank':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'].rank(pct=True))),
                    'cpt_salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                    'cpt_proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'] * 1.5)),
                    'cpt_own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['captain ownership']))
                }
            elif type_var == 'Showdown':
                if sport_var == 'NFL':
                    map_dict = {
                        'pos_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['position'])),
                        'team_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['team'])),
                        'salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                        'proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'])),
                        'own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'])),
                        'own_percent_rank':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'].rank(pct=True))),
                        'cpt_salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'] * 1.5)),
                        'cpt_proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'] * 1.5)),
                        'cpt_own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['captain ownership']))
                    }
                elif sport_var != 'NFL':
                    map_dict = {
                        'pos_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['position'])),
                        'team_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['team'])),
                        'salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'] / 1.5)),
                        'proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'])),
                        'own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'])),
                        'own_percent_rank':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'].rank(pct=True))),
                        'cpt_salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                        'cpt_proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'] * 1.5)),
                        'cpt_own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['captain ownership']))
                    }
        elif site_var == 'Fanduel':
            map_dict = {
                'pos_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['position'])),
                'team_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['team'])),
                'salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                'proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'])),
                'own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'])),
                'own_percent_rank':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['ownership'].rank(pct=True))),
                'cpt_salary_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['salary'])),
                'cpt_proj_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['median'] * 1.5)),
                'cpt_own_map':dict(zip(st.session_state['projections_df']['player_names'], st.session_state['projections_df']['captain ownership']))
            }
        
        if type_var == 'Classic':
            st.session_state['portfolio']['salary'] = st.session_state['portfolio'].apply(lambda row: sum(map_dict['salary_map'].get(player, 0) for player in row), axis=1)
            st.session_state['portfolio']['median'] = st.session_state['portfolio'].apply(lambda row: sum(map_dict['proj_map'].get(player, 0) for player in row), axis=1)
            st.session_state['portfolio']['Own'] = st.session_state['portfolio'].apply(lambda row: sum(map_dict['own_map'].get(player, 0) for player in row), axis=1)
            if stack_dict is not None:
                st.session_state['portfolio']['Stack'] = st.session_state['portfolio'].index.map(stack_dict)
        elif type_var == 'Showdown':
            # Calculate salary (CPT uses cpt_salary_map, others use salary_map)
            st.session_state['portfolio']['salary'] = st.session_state['portfolio'].apply(
                lambda row: map_dict['cpt_salary_map'].get(row.iloc[0], 0) + 
                          sum(map_dict['salary_map'].get(player, 0) for player in row.iloc[1:]),
                axis=1
            )
            
            # Calculate median (CPT uses cpt_proj_map, others use proj_map)
            st.session_state['portfolio']['median'] = st.session_state['portfolio'].apply(
                lambda row: map_dict['cpt_proj_map'].get(row.iloc[0], 0) + 
                          sum(map_dict['proj_map'].get(player, 0) for player in row.iloc[1:]),
                axis=1
            )
            
            # Calculate ownership (CPT uses cpt_own_map, others use own_map)
            st.session_state['portfolio']['Own'] = st.session_state['portfolio'].apply(
                lambda row: map_dict['cpt_own_map'].get(row.iloc[0], 0) + 
                          sum(map_dict['own_map'].get(player, 0) for player in row.iloc[1:]),
                axis=1
            )
        with col3:
            with st.form(key='filter_form'):
                max_dupes = st.number_input("Max acceptable dupes?", value=1000, min_value=1, step=1)
                min_salary = st.number_input("Min acceptable salary?", value=1000, min_value=1000, step=100)
                max_salary = st.number_input("Max acceptable salary?", value=60000, min_value=1000, step=100)
                max_finish_percentile = st.number_input("Max acceptable finish percentile?", value=.50, min_value=0.005, step=.001)
                player_names = set()
                for col in st.session_state['portfolio'].columns:
                    if col not in excluded_cols:
                        player_names.update(st.session_state['portfolio'][col].unique())
                player_lock = st.multiselect("Lock players?", options=sorted(list(player_names)), default=[])
                player_remove = st.multiselect("Remove players?", options=sorted(list(player_names)), default=[])
                if stack_dict is not None:
                    stack_toggle = st.selectbox("Include specific stacks?", options=['All Stacks', 'Specific Stacks'], index=0)
                    stack_selections = st.multiselect("If Specific Stacks, Which to include?", options=sorted(list(set(stack_dict.values()))), default=[])
                    stack_remove = st.multiselect("If Specific Stacks, Which to remove?", options=sorted(list(set(stack_dict.values()))), default=[])
                
                submitted = st.form_submit_button("Submit")

        with col2:
            st.session_state['portfolio'] = predict_dupes(st.session_state['portfolio'], map_dict, site_var, type_var, Contest_Size, strength_var)
            st.session_state['portfolio'] = st.session_state['portfolio'][st.session_state['portfolio']['Dupes'] <= max_dupes]
            st.session_state['portfolio'] = st.session_state['portfolio'][st.session_state['portfolio']['salary'] >= min_salary]
            st.session_state['portfolio'] = st.session_state['portfolio'][st.session_state['portfolio']['salary'] <= max_salary]
            st.session_state['portfolio'] = st.session_state['portfolio'][st.session_state['portfolio']['Finish_percentile'] <= max_finish_percentile]
            if stack_dict is not None:
                if stack_toggle == 'All Stacks':
                    st.session_state['portfolio'] = st.session_state['portfolio']
                    st.session_state['portfolio'] = st.session_state['portfolio'][~st.session_state['portfolio']['Stack'].isin(stack_remove)]
                else:
                    st.session_state['portfolio'] = st.session_state['portfolio'][st.session_state['portfolio']['Stack'].isin(stack_selections)]
                    st.session_state['portfolio'] = st.session_state['portfolio'][~st.session_state['portfolio']['Stack'].isin(stack_remove)]
            if player_remove:
                # Create mask for lineups that contain any of the removed players
                player_columns = [col for col in st.session_state['portfolio'].columns if col not in excluded_cols]
                remove_mask = st.session_state['portfolio'][player_columns].apply(
                    lambda row: not any(player in list(row) for player in player_remove), axis=1
                )
                st.session_state['portfolio'] = st.session_state['portfolio'][remove_mask]
            
            if player_lock:
                # Create mask for lineups that contain all locked players
                player_columns = [col for col in st.session_state['portfolio'].columns if col not in excluded_cols]
                
                lock_mask = st.session_state['portfolio'][player_columns].apply(
                    lambda row: all(player in list(row) for player in player_lock), axis=1
                )
                st.session_state['portfolio'] = st.session_state['portfolio'][lock_mask]
            export_file = st.session_state['portfolio'].copy()
            st.session_state['portfolio'] = st.session_state['portfolio'].sort_values(by='median', ascending=False)
            if csv_file is not None:
                player_columns = [col for col in st.session_state['portfolio'].columns if col not in excluded_cols]
                for col in player_columns:
                    export_file[col] = export_file[col].map(st.session_state['export_dict'])
            with st.expander("Download options"):
                if stack_dict is not None:
                    with st.form(key='stack_form'):
                        st.subheader("Stack Count Adjustments")
                        st.info("This allows you to fine tune the stacks that you wish to export. If you want to make sure you don't export any of a specific stack you can 0 it out.")
                        # Create a container for stack value inputs
                        sort_container = st.container()
                        with sort_container:
                            sort_var = st.selectbox("Sort export portfolio by:", options=['median', 'Lineup Edge', 'Own'])
                        
                        # Get unique stack values
                        unique_stacks = sorted(list(set(stack_dict.values())))
                        
                        # Create a dictionary to store stack multipliers
                        if 'stack_multipliers' not in st.session_state:
                            st.session_state.stack_multipliers = {stack: 0.0 for stack in unique_stacks}
                        
                        # Create columns for the stack inputs
                        num_cols = 6  # Number of columns to display
                        for i in range(0, len(unique_stacks), num_cols):
                            cols = st.columns(num_cols)
                            for j, stack in enumerate(unique_stacks[i:i+num_cols]):
                                with cols[j]:
                                    # Create a unique key for each number input
                                    key = f"stack_count_{stack}"
                                    # Get the current count of this stack in the portfolio
                                    current_stack_count = len(st.session_state['portfolio'][st.session_state['portfolio']['Stack'] == stack])
                                    # Create number input with current value and max value based on actual count
                                    st.session_state.stack_multipliers[stack] = st.number_input(
                                        f"{stack} count",
                                        min_value=0.0,
                                        max_value=float(current_stack_count),
                                        value=float(current_stack_count),
                                        step=1.0,
                                        key=key
                                    )
                        
                        # Create a copy of the portfolio
                        portfolio_copy = st.session_state['portfolio'].copy()
                        
                        # Create a list to store selected rows
                        selected_rows = []
                        
                        # For each stack, select the top N rows based on the count value
                        for stack in unique_stacks:
                            if stack in st.session_state.stack_multipliers:
                                count = int(st.session_state.stack_multipliers[stack])
                                # Get rows for this stack
                                stack_rows = portfolio_copy[portfolio_copy['Stack'] == stack]
                                # Sort by median and take top N rows
                                top_rows = stack_rows.nlargest(count, sort_var)
                                selected_rows.append(top_rows)
                        
                        # Combine all selected rows
                        portfolio_copy = pd.concat(selected_rows)
                        
                        # Update export_file with filtered data
                        export_file = portfolio_copy.copy()
                        
                        submitted = st.form_submit_button("Submit")
                        if submitted:
                            st.write('Export portfolio updated!')
                
            st.download_button(label="Download Portfolio", data=export_file.to_csv(index=False), file_name="portfolio.csv", mime="text/csv")
            # Display the paginated dataframe first
            st.dataframe(
                st.session_state['portfolio'].style
                .background_gradient(axis=0)
                .background_gradient(cmap='RdYlGn')
                .background_gradient(cmap='RdYlGn_r', subset=['Finish_percentile', 'Own', 'Dupes'])
                .format(freq_format, precision=2), 
                height=1000,
                use_container_width=True
            )

            # Add pagination controls below the dataframe
            total_rows = len(st.session_state['portfolio'])
            rows_per_page = 500
            total_pages = (total_rows + rows_per_page - 1) // rows_per_page  # Ceiling division

            # Initialize page number in session state if not exists
            if 'current_page' not in st.session_state:
                st.session_state.current_page = 1

            # Display current page range info and pagination control in a single line
            st.write(
                f"Showing rows {(st.session_state.current_page - 1) * rows_per_page + 1} "
                f"to {min(st.session_state.current_page * rows_per_page, total_rows)} of {total_rows}"
            )
            
            # Add page number input
            st.session_state.current_page = st.number_input(
                f"Page (1-{total_pages})", 
                min_value=1, 
                max_value=total_pages,
                value=st.session_state.current_page
            )

            # Calculate start and end indices for current page
            start_idx = (st.session_state.current_page - 1) * rows_per_page
            end_idx = min(start_idx + rows_per_page, total_rows)

            # Get the subset of data for the current page
            current_page_data = st.session_state['portfolio'].iloc[start_idx:end_idx]
            
            # Create player summary dataframe
            player_stats = []
            player_columns = [col for col in st.session_state['portfolio'].columns if col not in excluded_cols]
            
            if type_var == 'Showdown':
                # Handle Captain positions
                for player in player_names:
                    # Create mask for lineups where this player is Captain (first column)
                    cpt_mask = st.session_state['portfolio'][player_columns[0]] == player
                    
                    if cpt_mask.any():
                        player_stats.append({
                            'Player': f"{player} (CPT)",
                            'Lineup Count': cpt_mask.sum(),
                            'Avg Median': st.session_state['portfolio'][cpt_mask]['median'].mean(),
                            'Avg Own': st.session_state['portfolio'][cpt_mask]['Own'].mean(),
                            'Avg Dupes': st.session_state['portfolio'][cpt_mask]['Dupes'].mean(),
                            'Avg Finish %': st.session_state['portfolio'][cpt_mask]['Finish_percentile'].mean(),
                            'Avg Lineup Edge': st.session_state['portfolio'][cpt_mask]['Lineup Edge'].mean(),
                        })
                    
                    # Create mask for lineups where this player is FLEX (other columns)
                    flex_mask = st.session_state['portfolio'][player_columns[1:]].apply(
                        lambda row: player in list(row), axis=1
                    )
                    
                    if flex_mask.any():
                        player_stats.append({
                            'Player': f"{player} (FLEX)",
                            'Lineup Count': flex_mask.sum(),
                            'Avg Median': st.session_state['portfolio'][flex_mask]['median'].mean(),
                            'Avg Own': st.session_state['portfolio'][flex_mask]['Own'].mean(),
                            'Avg Dupes': st.session_state['portfolio'][flex_mask]['Dupes'].mean(),
                            'Avg Finish %': st.session_state['portfolio'][flex_mask]['Finish_percentile'].mean(),
                            'Avg Lineup Edge': st.session_state['portfolio'][flex_mask]['Lineup Edge'].mean(),
                        })
            else:
                # Original Classic format processing
                for player in player_names:
                    player_mask = st.session_state['portfolio'][player_columns].apply(
                        lambda row: player in list(row), axis=1
                    )
                    
                    if player_mask.any():
                        player_stats.append({
                            'Player': player,
                            'Lineup Count': player_mask.sum(),
                            'Avg Median': st.session_state['portfolio'][player_mask]['median'].mean(),
                            'Avg Own': st.session_state['portfolio'][player_mask]['Own'].mean(),
                            'Avg Dupes': st.session_state['portfolio'][player_mask]['Dupes'].mean(),
                            'Avg Finish %': st.session_state['portfolio'][player_mask]['Finish_percentile'].mean(),
                            'Avg Lineup Edge': st.session_state['portfolio'][player_mask]['Lineup Edge'].mean(),
                        })
            
            player_summary = pd.DataFrame(player_stats)
            player_summary = player_summary.sort_values('Lineup Count', ascending=False)
            
            st.subheader("Player Summary")
            st.dataframe(
                player_summary.style
                .background_gradient(axis=0).background_gradient(cmap='RdYlGn').background_gradient(cmap='RdYlGn_r', subset=['Avg Finish %', 'Avg Own', 'Avg Dupes'])
                .format({
                    'Avg Median': '{:.2f}',
                    'Avg Own': '{:.2f}',
                    'Avg Dupes': '{:.2f}',
                    'Avg Finish %': '{:.2%}',
                    'Avg Lineup Edge': '{:.2%}'
                }),
                height=400,
                use_container_width=True
            )