Analyzing Simulation Results

Learn how to extract insights from simulation data.

Database Connection

Connect to the SQLite database:

import sqlite3
import pandas as pd

conn = sqlite3.connect("simulation.db")

Basic Queries

Action Distribution

SELECT action_type, COUNT(*) as count,
       SUM(CASE WHEN success = 1 THEN 1 ELSE 0 END) as successful,
       AVG(duration) as avg_duration
FROM action_logs
GROUP BY action_type
ORDER BY count DESC;

Combat Statistics

SELECT attacker_id,
       COUNT(*) as attacks,
       SUM(damage_dealt) as total_damage,
       AVG(damage_dealt) as avg_damage,
       SUM(CASE WHEN target_died = 1 THEN 1 ELSE 0 END) as kills
FROM combat_logs
GROUP BY attacker_id
ORDER BY kills DESC;

Trading Patterns

SELECT initiator_id, target_id,
       COUNT(*) as trades,
       offered_items,
       requested_items
FROM trade_logs
WHERE completed = 1
GROUP BY initiator_id, target_id;

Python Analysis

Using Pandas

import pandas as pd

# Load action logs
actions = pd.read_sql_query("SELECT * FROM action_logs", conn)

# Success rate by action type
success_rate = actions.groupby('action_type')['success'].mean()
print(success_rate.sort_values(ascending=False))

# Actions over time
actions_per_tick = actions.groupby('tick').size()
actions_per_tick.plot(title='Activity Over Time')

Agent Performance

# Get final agent states
final_tick = pd.read_sql_query(
    "SELECT MAX(tick) as max_tick FROM agent_snapshots",
    conn
).iloc[0]['max_tick']

final_agents = pd.read_sql_query(
    f"SELECT * FROM agent_snapshots WHERE tick = {final_tick}",
    conn
)

# Agents by health
print(final_agents[['name', 'health', 'max_health']].sort_values('health'))

# Inventory sizes
print(final_agents[['name', 'inventory_items']].sort_values('inventory_items', ascending=False))

Economic Analysis

Resource Gathering Rates

gather_actions = pd.read_sql_query("""
    SELECT agent_id, tick, action_data
    FROM action_logs
    WHERE action_type LIKE '%Gather%' AND success = 1
""", conn)

# Parse JSON action_data
import json
gather_actions['resource'] = gather_actions['action_data'].apply(
    lambda x: json.loads(x).get('resource_type')
)

# Resources per agent
resources_gathered = gather_actions.groupby(['agent_id', 'resource']).size()
print(resources_gathered.unstack(fill_value=0))

Trade Network Visualization

import networkx as nx
import matplotlib.pyplot as plt

trades = pd.read_sql_query(
    "SELECT initiator_id, target_id FROM trade_logs WHERE completed = 1",
    conn
)

# Create trade network
G = nx.from_pandas_edgelist(
    trades,
    source='initiator_id',
    target='target_id',
    create_using=nx.DiGraph()
)

# Visualize
nx.draw(G, with_labels=True, node_color='lightblue', node_size=500)
plt.title("Trade Network")
plt.show()

Time Series Analysis

Agent Health Over Time

agent_health = pd.read_sql_query("""
    SELECT tick, agent_id, name, health, max_health
    FROM agent_snapshots
    ORDER BY tick
""", conn)

# Plot health trajectories
for agent_id in agent_health['agent_id'].unique()[:5]:
    agent_data = agent_health[agent_health['agent_id'] == agent_id]
    plt.plot(agent_data['tick'], agent_data['health'],
             label=agent_data['name'].iloc[0])

plt.xlabel('Tick')
plt.ylabel('Health')
plt.legend()
plt.title('Agent Health Over Time')
plt.show()

Exploration Progress

# Count unique positions visited per agent
snapshots = pd.read_sql_query("""
    SELECT tick, agent_id, position_x, position_y
    FROM agent_snapshots
""", conn)

snapshots['position'] = list(zip(snapshots['position_x'], snapshots['position_y']))

exploration = snapshots.groupby(['agent_id', 'tick'])['position'].nunique()
exploration.unstack(level=0).plot(title='Exploration Over Time')

Advanced Analytics

Using the built-in analytics script:

python analyze_simulation_results.py simulation.db

This generates:

Summary statistics
Action distribution charts
Combat analysis
Trade networks
Resource flow diagrams
Agent performance rankings

Export Results

Export for external analysis:

# Export to CSV
actions.to_csv('actions.csv', index=False)
final_agents.to_csv('final_agents.csv', index=False)

# Export to Excel
with pd.ExcelWriter('simulation_results.xlsx') as writer:
    actions.to_excel(writer, sheet_name='Actions')
    final_agents.to_excel(writer, sheet_name='Agents')
    trades.to_excel(writer, sheet_name='Trades')

Next Steps

Experiment with different agent configurations
Compare multiple simulation runs
Publish findings from your research