Backtesting with Fivesight

Backtesting allows you to simulate trading strategies against historical Polymarket data using a Virtual Clock.

Prerequisites

• A running fivesight-backend.
• fivesight-clob-client installed.

Step-by-Step Guide

1. Initialize Client & Session

from fivesight_clob_client.client import ClobClient
from datetime import datetime, timedelta

client = ClobClient(
    host="https://api.fivesight.ai/api/simulation", 
    key="dummy", 
    chain_id=137, 
    simulation_mode=True
)

# Create BACKTEST session
client.create_simulation_session(mode="BACKTEST", initial_balance=50000)

2. Set the Virtual Clock

You must manually advance the clock. The simulation engine will return the state of the market (Order Book) as it was at that exact timestamp.

start_time = datetime(2023, 12, 28, 12, 0, 0)
client.set_simulation_clock(start_time.isoformat())

3. Fetch Historical Data

When you call get_order_book, the backend fetches the snapshot closest to your Virtual Clock.

token_id = "1234..." # Valid Token ID
book = client.get_order_book(token_id)
print(book)

4. Backtest Loop

current_time = start_time
for _ in range(60): # Simulate 1 hour
    # 1. Update Clock
    client.set_simulation_clock(current_time.isoformat())
    
    # 2. Get Data
    book = client.get_order_book(token_id)
    
    # 3. Execute Strategy
    # ... logic to buy/sell ...
    # client.create_and_post_order(...)
    
    # 4. Advance Time
    current_time += timedelta(minutes=1)

Supported Features

• Historical Order Book: Fetched from secondary pm_data backend.
• Immediate Matching: Orders crossing the spread are executed immediately against the historical snapshot.
• Position Tracking: Your SimulatedPosition and SimulatedWallet are updated automatically.