# Define the date range for the last year
end_date = datetime.now()
start_date = end_date - timedelta(days=365)

# Team composition pipeline to get the top 5 teams based on total matches
top_teams_pipeline = [
    {
        "$unwind": "$players.team1"
    },
    {
        "$match": {
            "date": {
                "$gte": int(start_date.timestamp() * 1000),
                "$lte": int(end_date.timestamp() * 1000)
            }
        }
    },
    {
        "$group": {
            "_id": "$team1.name",
            "totalMatches": { "$sum": 1 }
        }
    },
    {
        "$sort": {
            "totalMatches": -1
        }
    },
    {
        "$limit": 5
    }
]

top_teams_results = list(collection.aggregate(top_teams_pipeline))
top_teams = [team['_id'] for team in top_teams_results]

# Roster changes pipeline limited to the top 5 teams and the last year
roster_changes_pipeline = [
    {
        "$match": {
            "date": {
                "$gte": int(start_date.timestamp() * 1000),
                "$lte": int(end_date.timestamp() * 1000)
            },
            "team1.name": { "$in": top_teams }
        }
    },
    {
        "$unwind": "$players.team1"
    },
    {
        "$group": {
            "_id": {
                "team": "$team1.name",
                "date": "$date"
            },
            "players": { "$addToSet": "$players.team1.name" }
        }
    },
    {
        "$group": {
            "_id": "$_id.team",
            "rosterChanges": {
                "$push": {
                    "date": "$_id.date",
                    "players": "$players"
                }
            }
        }
    }
]

roster_changes_results = list(collection.aggregate(roster_changes_pipeline))

# Simulate a calculation of roster changes over time
roster_changes = []
for team in roster_changes_results:
    for i in range(1, len(team['rosterChanges'])):
        if team['rosterChanges'][i]['players'] != team['rosterChanges'][i-1]['players']:
            roster_changes.append({
                'team': team['_id'],
                'date': team['rosterChanges'][i]['date'],
                'roster_change': True
            })
        else:
            roster_changes.append({
                'team': team['_id'],
                'date': team['rosterChanges'][i]['date'],
                'roster_change': False
            })

roster_changes_df = pd.DataFrame(roster_changes)

# Performance data pipeline limited to the top 5 teams and the last year
performance_pipeline = [
    {
        "$match": {
            "date": {
                "$gte": int(start_date.timestamp() * 1000),
                "$lte": int(end_date.timestamp() * 1000)
            },
            "team1.name": { "$in": top_teams }
        }
    },
    {
        "$group": {
            "_id": {
                "team": "$team1.name",
                "date": "$date"
            },
            "totalMatches": { "$sum": 1 },
            "wins": {
                "$sum": {
                    "$cond": [{ "$eq": ["$winnerTeam.name", "$team1.name"] }, 1, 0]
                }
            }
        }
    },
    {
        "$project": {
            "_id": 0,
            "team": "$_id.team",
            "date": "$_id.date",
            "totalMatches": 1,
            "wins": 1,
            "winRate": {
                "$multiply": [
                    { "$divide": ["$wins", "$totalMatches"] },
                    100
                ]
            }
        }
    }
]

performance_results = list(collection.aggregate(performance_pipeline))
performance_df = pd.DataFrame(performance_results)

# Merge the roster changes DataFrame with the performance DataFrame
merged_df = pd.merge(roster_changes_df, performance_df, on=['team', 'date'])

# Convert timestamp to datetime for plotting
merged_df['date'] = pd.to_datetime(merged_df['date'], unit='ms')

# Plot the impact of roster changes on team performance
plt.figure(figsize=(12, 8))
sns.lineplot(x='date', y='winRate', hue='roster_change', style='team', data=merged_df, markers=True)
plt.title('Impact of Roster Changes on Team Dynamics (Top 5 Teams, Last Year)')
plt.xlabel('Date')
plt.ylabel('Win Rate (%)')
plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left', title='Roster Change')
plt.tight_layout()
plt.show()