app.py

import streamlit as st
st.set_page_config(layout="wide")
import numpy as np
import pandas as pd
import pymongo
import time
from datetime import datetime, timedelta
from scipy import stats

@st.cache_resource
def init_conn():
        
        uri = st.secrets['mongo_uri']
        client = pymongo.MongoClient(uri, retryWrites=True, serverSelectionTimeoutMS=500000)
        db = client["League_of_Legends_Database"]

        current_date = datetime.now()

        collection = db["gamelogs"] 
        max_date = current_date - timedelta(days=1)
        min_date = current_date - timedelta(days=365)
        team_names = collection.distinct("teamname")
        player_names = collection.distinct("playername")

        return db, team_names, player_names, min_date, max_date
    
db, team_names, player_names, min_date, max_date = init_conn()

display_formats = {'wKill%': '{:.2%}', 'wDeath%': '{:.2%}', 'wAssist%': '{:.2%}', 'lKill%': '{:.2%}', 'lDeath%': '{:.2%}', 'lAssist%': '{:.2%}'}

# Create sidebar container for options
with st.sidebar:
    st.header("Team Analysis Options")

    # Date filtering options
    st.subheader("Date Range")
    date_filter = st.radio(
        "Select Date Range",
        ["Last Year", "Custom Range"]
    )
    
    if date_filter == "Last Year":
        end_date = max_date
        start_date = (end_date - timedelta(days=365))
    else:
        col1, col2 = st.columns(2)
        with col1:
            start_date = st.date_input(
                "Start Date",
                value=max_date.date() - timedelta(days=30),
                min_value=min_date.date(),
                max_value=max_date.date()
            )
        with col2:
            end_date = st.date_input(
                "End Date",
                value=max_date.date(),
                min_value=min_date.date(),
                max_value=max_date.date()
            )

    col1, col2 = st.columns(2)
    with col1:
        selected_team = st.selectbox(
            "Select Team",
            options=team_names,
            index=team_names.index("T1") if "T1" in team_names else 0
        )
    with col2:
        selected_opponent = st.selectbox(
            "Select Opponent",
            options=team_names,
            index=team_names.index("T1") if "T1" in team_names else 0
        )

    st.subheader("Prediction Settings")
    num_games = st.selectbox(
        "Is the match BO1, BO3, or BO5?",
        options=["BO1", "BO3", "BO5"],
        index=0
    )
    
    # Convert BO format to number of games
    game_count = int(num_games[2])
    
    # Create lists to store settings for each game
    win_loss_settings = []
    game_settings_list = []
    kill_predictions = []
    death_predictions = []
    
    # Create a tab for each game
    game_tabs = st.tabs([f"Game {i+1}" for i in range(game_count)])
    
    for game_num, game_tab in enumerate(game_tabs, 1):
        with game_tab:

            win_loss_settings.append(st.selectbox(
                f"Game {game_num} Win/Loss",
                options=["Win", "Loss"],
                index=0,
                key=f"win_loss_{game_num}"
            ))
            game_setting = st.selectbox(
                f"Game {game_num} Prediction Type",
                options=["Average", "Predict"],
                index=0,
                key=f"game_settings_{game_num}"
            )

            if game_setting == "Average":
                kill_predictions.append(0)
                death_predictions.append(0)
            else:
                col1, col2 = st.columns(2)
                with col1:
                    kill_predictions.append(st.number_input(
                        f"Game {game_num} Predicted Team Kills",
                        min_value=1,
                        max_value=100,
                        value=20,
                        key=f"kills_{game_num}"
                    ))
                with col2:
                    death_predictions.append(st.number_input(
                        f"Game {game_num} Predicted Team Deaths",
                        min_value=1,
                        max_value=100,
                        value=5,
                        key=f"deaths_{game_num}"
                    ))

@st.cache_data()
def simulate_stats(row, num_sims=1000):
    """Simulate stats using normal distribution"""
    # Using coefficient of variation of 0.3 to generate reasonable standard deviations
    cv = 0.3
    percentiles = [10, 25, 50, 75, 90]
    
    results = {}
    for stat in ['Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']:
        mean = row[stat]
        std = mean * cv  # Using coefficient of variation to determine std
        sims = stats.norm.rvs(loc=mean, scale=std, size=num_sims)
        # Ensure no negative values
        sims = np.maximum(sims, 0)
        results[stat] = np.percentile(sims, percentiles)
    
    return pd.Series(results)

@st.cache_data(ttl = 60)
def init_team_data(team, opponent, win_loss_settings, kill_predictions, death_predictions, start_date, end_date):  
        game_count = len(kill_predictions)
        overall_team_data = pd.DataFrame(columns = ['playername', 'teamname', 'position', 'Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj'])
        # Convert date objects to datetime strings in the correct format
        start_datetime = datetime.combine(start_date, datetime.min.time()).strftime("%Y-%m-%d %H:%M:%S")
        end_datetime = datetime.combine(end_date, datetime.max.time()).strftime("%Y-%m-%d %H:%M:%S")

        collection = db["gamelogs"] 
        cursor = collection.find({"teamname": team, "date": {"$gte": start_datetime, "$lte": end_datetime}})
        raw_display = pd.DataFrame(list(cursor))

        cursor = collection.find({"date": {"$gte": start_datetime, "$lte": end_datetime}})
        raw_opponent = pd.DataFrame(list(cursor))

        tables_to_loop = [raw_display, raw_opponent]

        for loop in range(len(tables_to_loop)):
            tables = tables_to_loop[loop]
            calc_columns = ['kills', 'deaths', 'assists', 'total_cs']
            league_pos_win_stats = {}   
            league_pos_loss_stats = {}
            Opponent_pos_win_allowed_stats = {}
            Opponent_pos_loss_allowed_stats = {}
            playername_win_stats = {}
            playername_loss_stats = {}
            teamname_win_stats = {}
            teamname_loss_stats = {}

            if loop == 0:

                for stats in calc_columns:
                    playername_win_stats[stats] = tables[tables['result'] == 1].groupby(['playername'])[stats].mean().to_dict()
                    playername_loss_stats[stats] = tables[tables['result'] == 0].groupby(['playername'])[stats].mean().to_dict()
                    teamname_win_stats[stats] = tables[(tables['result'] == 1) & (tables['position'] == 'team')].groupby(['teamname'])[stats].mean().to_dict()
                    teamname_loss_stats[stats] = tables[(tables['result'] == 0) & (tables['position'] == 'team')].groupby(['teamname'])[stats].mean().to_dict() 

                for stat in calc_columns:

                    column_name = f'playername_avg_{stat}_win'
                    tables[column_name] = tables.apply(
                        lambda row: playername_win_stats[stat].get(row['playername'], 0), 
                        axis=1
                    )

                    column_name = f'playername_avg_{stat}_loss'
                    tables[column_name] = tables.apply(
                        lambda row: playername_loss_stats[stat].get(row['playername'], 0), 
                        axis=1
                    )

                    column_name = f'teamname_avg_{stat}_win'
                    tables[column_name] = tables.apply(
                        lambda row: teamname_win_stats[stat].get(row['teamname'], 0), 
                        axis=1
                    )

                    column_name = f'teamname_avg_{stat}_loss'
                    tables[column_name] = tables.apply(
                        lambda row: teamname_loss_stats[stat].get(row['teamname'], 0), 
                        axis=1
                    )
            
                tables['playername_avg_kill_share_win'] = tables['playername_avg_kills_win'] / tables['teamname_avg_kills_win']
                tables['playername_avg_death_share_win'] = tables['playername_avg_deaths_win'] / tables['teamname_avg_deaths_win']
                tables['playername_avg_assist_share_win'] = tables['playername_avg_assists_win'] / tables['teamname_avg_kills_win']
                tables['playername_avg_cs_share_win'] = tables['playername_avg_total_cs_win'] / tables['teamname_avg_total_cs_win']
                tables['playername_avg_kill_share_loss'] = tables['playername_avg_kills_loss'] / tables['teamname_avg_kills_loss']
                tables['playername_avg_death_share_loss'] = tables['playername_avg_deaths_loss'] / tables['teamname_avg_deaths_loss']
                tables['playername_avg_assist_share_loss'] = tables['playername_avg_assists_loss'] / tables['teamname_avg_kills_loss']
                tables['playername_avg_cs_share_loss'] = tables['playername_avg_total_cs_loss'] / tables['teamname_avg_total_cs_loss']
                player_tables = tables

            else:

                for stats in calc_columns:
                    league_pos_win_stats[stats] = {
                        league: group.groupby('position')[stats].mean().to_dict()
                        for league, group in tables[tables['result'] == 1].groupby('league')
                    }
                    league_pos_loss_stats[stats] = {
                        league: group.groupby('position')[stats].mean().to_dict()
                        for league, group in tables[tables['result'] == 0].groupby('league')
                    }
                    
                    Opponent_pos_win_allowed_stats[stats] = {
                        opponent: group.groupby('position')[stats].mean().to_dict()
                        for opponent, group in tables[tables['result'] == 1].groupby('Opponent')
                    }
                    Opponent_pos_loss_allowed_stats[stats] = {
                        opponent: group.groupby('position')[stats].mean().to_dict()
                        for opponent, group in tables[tables['result'] == 0].groupby('Opponent')
                    }

                for stat in calc_columns:

                    column_name = f'league_pos_avg_{stat}_win'
                    tables[column_name] = tables.apply(
                        lambda row: league_pos_win_stats[stat].get(row['league'], {}).get(row['position'], 0), 
                        axis=1
                    )
                    
                    column_name = f'league_pos_avg_{stat}_loss'
                    tables[column_name] = tables.apply(
                        lambda row: league_pos_loss_stats[stat].get(row['league'], {}).get(row['position'], 0), 
                        axis=1
                    )

                    column_name = f'Opponent_pos_avg_{stat}_allowed_win'
                    tables[column_name] = tables.apply(
                        lambda row: Opponent_pos_win_allowed_stats[stat].get(row['Opponent'], {}).get(row['position'], 0), 
                        axis=1
                    )

                    column_name = f'Opponent_pos_avg_{stat}_allowed_loss'
                    tables[column_name] = tables.apply(
                        lambda row: Opponent_pos_loss_allowed_stats[stat].get(row['Opponent'], {}).get(row['position'], 0), 
                        axis=1
                    )
                
                tables = tables[tables['Opponent'] == opponent]

                tables['overall_win_kills_boost_pos'] = tables['Opponent_pos_avg_kills_allowed_win'] / tables['league_pos_avg_kills_win']
                tables['overall_win_deaths_boost_pos'] = tables['Opponent_pos_avg_deaths_allowed_win'] / tables['league_pos_avg_deaths_win']
                tables['overall_win_assists_boost_pos'] = tables['Opponent_pos_avg_assists_allowed_win'] / tables['league_pos_avg_assists_win']
                tables['overall_win_total_cs_boost_pos'] = tables['Opponent_pos_avg_total_cs_allowed_win'] / tables['league_pos_avg_total_cs_win']
                tables['overall_loss_kills_boost_pos'] = tables['Opponent_pos_avg_kills_allowed_loss'] / tables['league_pos_avg_kills_loss']
                tables['overall_loss_deaths_boost_pos'] = tables['Opponent_pos_avg_deaths_allowed_loss'] / tables['league_pos_avg_deaths_loss']
                tables['overall_loss_assists_boost_pos'] = tables['Opponent_pos_avg_assists_allowed_loss'] / tables['league_pos_avg_assists_loss']
                tables['overall_loss_total_cs_boost_pos'] = tables['Opponent_pos_avg_total_cs_allowed_loss'] / tables['league_pos_avg_total_cs_loss']
                
                opp_tables = tables
                opp_pos_kills_boost_win = dict(zip(opp_tables['position'], opp_tables['overall_win_kills_boost_pos']))
                opp_pos_deaths_boost_win = dict(zip(opp_tables['position'], opp_tables['overall_win_deaths_boost_pos']))
                opp_pos_assists_boost_win = dict(zip(opp_tables['position'], opp_tables['overall_win_assists_boost_pos']))
                opp_pos_cs_boost_win = dict(zip(opp_tables['position'], opp_tables['overall_win_total_cs_boost_pos']))
                opp_pos_kills_boost_loss = dict(zip(opp_tables['position'], opp_tables['overall_loss_kills_boost_pos']))
                opp_pos_deaths_boost_loss = dict(zip(opp_tables['position'], opp_tables['overall_loss_deaths_boost_pos']))
                opp_pos_assists_boost_loss = dict(zip(opp_tables['position'], opp_tables['overall_loss_assists_boost_pos']))
                opp_pos_cs_boost_loss = dict(zip(opp_tables['position'], opp_tables['overall_loss_total_cs_boost_pos']))
                opp_boosts = pd.DataFrame({
                    'opp_pos_kills_boost_win': opp_pos_kills_boost_win,
                    'opp_pos_deaths_boost_win': opp_pos_deaths_boost_win,
                    'opp_pos_assists_boost_win': opp_pos_assists_boost_win,
                    'opp_pos_cs_boost_win': opp_pos_cs_boost_win,
                    'opp_pos_kills_boost_loss': opp_pos_kills_boost_loss,
                    'opp_pos_deaths_boost_loss': opp_pos_deaths_boost_loss,
                    'opp_pos_assists_boost_loss': opp_pos_assists_boost_loss,
                    'opp_pos_cs_boost_loss': opp_pos_cs_boost_loss
                }).set_index(pd.Index(list(opp_pos_kills_boost_win.keys()), name='position'))

        results_dict = {}

        for game in range(game_count):
            if kill_predictions[game] > 0:
                working_tables = player_tables[['playername', 'teamname', 'position', 'playername_avg_kill_share_win', 'playername_avg_death_share_win','playername_avg_assist_share_win',
                                        'playername_avg_total_cs_win', 'playername_avg_kill_share_loss', 'playername_avg_death_share_loss', 'playername_avg_assist_share_loss', 'playername_avg_total_cs_loss']]
                working_tables = working_tables.rename(columns = {'playername_avg_kill_share_win': 'wKill%', 'playername_avg_death_share_win': 'wDeath%', 'playername_avg_assist_share_win': 'wAssist%',
                                                            'playername_avg_total_cs_win': 'wCS', 'playername_avg_kill_share_loss': 'lKill%', 'playername_avg_death_share_loss': 'lDeath%',
                                                            'playername_avg_assist_share_loss': 'lAssist%', 'playername_avg_total_cs_loss': 'lCS'})
                team_data = working_tables.drop_duplicates(subset = ['playername'])

                if win_loss_settings[game] == "Win":
                    team_data['Kill_Proj'] = team_data.apply(lambda row: row['wKill%'] * opp_pos_kills_boost_win.get(row['position'], 1), axis=1) * kill_predictions[game]
                    team_data['Death_Proj'] = team_data.apply(lambda row: row['wDeath%'] * opp_pos_deaths_boost_win.get(row['position'], 1), axis=1) * death_predictions[game]
                    team_data['Assist_Proj'] = team_data.apply(lambda row: row['wAssist%'] * opp_pos_assists_boost_win.get(row['position'], 1), axis=1) * kill_predictions[game]
                    team_data['CS_Proj'] = team_data.apply(lambda row: row['wCS'] * opp_pos_cs_boost_win.get(row['position'], 1), axis=1)
                    team_data = team_data[['playername', 'teamname', 'position', 'Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']]
                else:
                    team_data['Kill_Proj'] = team_data.apply(lambda row: row['lKill%'] * opp_pos_kills_boost_loss.get(row['position'], 1), axis=1) * kill_predictions[game]
                    team_data['Death_Proj'] = team_data.apply(lambda row: row['lDeath%'] * opp_pos_deaths_boost_loss.get(row['position'], 1), axis=1) * death_predictions[game]
                    team_data['Assist_Proj'] = team_data.apply(lambda row: row['lAssist%'] * opp_pos_assists_boost_loss.get(row['position'], 1), axis=1) * kill_predictions[game]
                    team_data['CS_Proj'] = team_data.apply(lambda row: row['lCS'] * opp_pos_cs_boost_loss.get(row['position'], 1), axis=1)
                    team_data = team_data[['playername', 'teamname', 'position', 'Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']]
            else:
                working_tables = player_tables[['playername', 'teamname', 'position', 'playername_avg_kills_win', 'playername_avg_deaths_win', 'playername_avg_assists_win', 'playername_avg_total_cs_win',
                                        'playername_avg_kills_loss', 'playername_avg_deaths_loss', 'playername_avg_assists_loss', 'playername_avg_total_cs_loss']]
                working_tables = working_tables.rename(columns = {'playername_avg_kills_win': 'wKill%', 'playername_avg_deaths_win': 'wDeath%', 'playername_avg_assists_win': 'wAssist%',
                                                            'playername_avg_total_cs_win': 'wCS', 'playername_avg_kills_loss': 'lKill%', 'playername_avg_deaths_loss': 'lDeath%',
                                                            'playername_avg_assists_loss': 'lAssist%', 'playername_avg_total_cs_loss': 'lCS'})
                team_data = working_tables.drop_duplicates(subset = ['playername'])

                if win_loss_settings[game] == "Win":
                    team_data['Kill_Proj'] = team_data.apply(lambda row: row['wKill%'] * opp_pos_kills_boost_win.get(row['position'], 1), axis=1)
                    team_data['Death_Proj'] = team_data.apply(lambda row: row['wDeath%'] * opp_pos_deaths_boost_win.get(row['position'], 1), axis=1)
                    team_data['Assist_Proj'] = team_data.apply(lambda row: row['wAssist%'] * opp_pos_assists_boost_win.get(row['position'], 1), axis=1)
                    team_data['CS_Proj'] = team_data.apply(lambda row: row['wCS'] * opp_pos_cs_boost_win.get(row['position'], 1), axis=1)
                    team_data = team_data[['playername', 'teamname', 'position', 'Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']]
                else:
                    team_data['Kill_Proj'] = team_data.apply(lambda row: row['lKill%'] * opp_pos_kills_boost_loss.get(row['position'], 1), axis=1)
                    team_data['Death_Proj'] = team_data.apply(lambda row: row['lDeath%'] * opp_pos_deaths_boost_loss.get(row['position'], 1), axis=1)
                    team_data['Assist_Proj'] = team_data.apply(lambda row: row['lAssist%'] * opp_pos_assists_boost_loss.get(row['position'], 1), axis=1)
                    team_data['CS_Proj'] = team_data.apply(lambda row: row['lCS'] * opp_pos_cs_boost_loss.get(row['position'], 1), axis=1)
                    team_data = team_data[['playername', 'teamname', 'position', 'Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']]
            
            results_dict[f'game {game + 1}'] = team_data.dropna()
            team_data['playername'] = team_data['playername'] + f' game {game + 1}'

            overall_team_data = pd.concat([overall_team_data, team_data])

        return overall_team_data.dropna().set_index('playername'), opp_boosts, results_dict

if st.button("Run"):
    team_data, opp_boost, results_dict = init_team_data(selected_team, selected_opponent, win_loss_settings, kill_predictions, death_predictions, start_date, end_date)

    player_summary = pd.DataFrame()
    for game_num, game_df in results_dict.items():
        # Remove 'game X' from playernames if present
        clean_df = game_df.copy()
        clean_df['playername'] = clean_df['playername'].str.split(' game ').str[0]
        
        if player_summary.empty:
            player_summary = clean_df
        else:
            # Add the stats to existing players
            player_summary.update(clean_df)  # Update teamname and position if needed
            for col in ['Kill_Proj', 'Death_Proj', 'Assist_Proj', 'CS_Proj']:
                player_summary[col] += clean_df[col]
    player_summary = player_summary.set_index('playername')

    # Create simulated percentiles
    individual_sim_results = []
    for idx, row in team_data.iterrows():
        percentiles = simulate_stats(row)
        individual_sim_results.append({
            'Player': idx,
            'Position': row['position'],
            'Stat': 'Kills',
            '10%': percentiles['Kill_Proj'][0],
            '25%': percentiles['Kill_Proj'][1],
            '50%': percentiles['Kill_Proj'][2],
            '75%': percentiles['Kill_Proj'][3],
            '90%': percentiles['Kill_Proj'][4]
        })
        # Repeat for other stats
        for stat, name in [('Death_Proj', 'Deaths'), ('Assist_Proj', 'Assists'), ('CS_Proj', 'CS')]:
            individual_sim_results.append({
                'Player': idx,
                'Position': row['position'],
                'Stat': name,
                '10%': percentiles[stat][0],
                '25%': percentiles[stat][1],
                '50%': percentiles[stat][2],
                '75%': percentiles[stat][3],
                '90%': percentiles[stat][4]
            })
    
    sim_df = pd.DataFrame(individual_sim_results)

    # Create simulated percentiles
    overall_sim_results = []
    for idx, row in player_summary.iterrows():
        percentiles = simulate_stats(row)
        overall_sim_results.append({
            'Player': idx,
            'Position': row['position'],
            'Stat': 'Kills',
            '10%': percentiles['Kill_Proj'][0],
            '25%': percentiles['Kill_Proj'][1],
            '50%': percentiles['Kill_Proj'][2],
            '75%': percentiles['Kill_Proj'][3],
            '90%': percentiles['Kill_Proj'][4]
        })
        # Repeat for other stats
        for stat, name in [('Death_Proj', 'Deaths'), ('Assist_Proj', 'Assists'), ('CS_Proj', 'CS')]:
            overall_sim_results.append({
                'Player': idx,
                'Position': row['position'],
                'Stat': name,
                '10%': percentiles[stat][0],
                '25%': percentiles[stat][1],
                '50%': percentiles[stat][2],
                '75%': percentiles[stat][3],
                '90%': percentiles[stat][4]
            })
    
    overall_sim_df = pd.DataFrame(overall_sim_results)
    overall_sim_df = overall_sim_df.drop_duplicates(subset = ['Player', 'Stat'])

    tab1, tab2, tab3 = st.tabs(["Overall Data", "Individual Game Data", "Opponent Data"])
    with tab1:
        st.subheader("Full Match Data")
        st.dataframe(player_summary.style.background_gradient(axis=0).background_gradient(cmap='RdYlGn').format(display_formats, precision=2), use_container_width = True)

        st.subheader("Overall Simulations")
        stat_tabs = st.tabs(["Kills", "Deaths", "Assists", "CS"])
        
        for stat, tab in zip(["Kills", "Deaths", "Assists", "CS"], stat_tabs):
            with tab:
                stat_data = overall_sim_df[overall_sim_df['Stat'] == stat].copy()
                stat_data = stat_data.set_index('Player')[['Position', '10%', '25%', '50%', '75%', '90%']]
                st.dataframe(
                    stat_data.style.format(precision=2).background_gradient(axis=0).background_gradient(cmap='RdYlGn'),
                    use_container_width=True
                )

    with tab2:
        st.subheader("Individual Game Data")
        st.dataframe(team_data.style.background_gradient(axis=0).background_gradient(cmap='RdYlGn').format(display_formats, precision=2), use_container_width = True)

        st.subheader("Individual Game Simulations")
        unique_players = sim_df['Player'].unique().tolist()
        player_tabs = st.tabs(unique_players)
        
        for player, tab in zip(unique_players, player_tabs):
            with tab:
                player_data = sim_df[sim_df['Player'] == player]
                player_data = player_data.set_index('Stat')
                st.dataframe(
                    player_data[['10%', '25%', '50%', '75%', '90%']]
                    .style.format(precision=2),
                    use_container_width=True
                )
                
    with tab3:
        st.dataframe(opp_boost.style.background_gradient(axis=0).background_gradient(cmap='RdYlGn').format(precision=2), use_container_width = True)